MIRIAD Users Guide for SMA

Transcription

1 MIRIAD Users Guide for SMA Jun-Hui Zhao July 9, 2012

2 iriad-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <M -for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> i <Miriad- MA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-S Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> < d-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miria -SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for > <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA iriad-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Mi for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-f A> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SM Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> < d-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miria -SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for > <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA iriad-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Mi for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-f A> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SM Miriad-for-SMA> <Miriad-for-SMA> Multichannel Image <Miriad-for-SMA> Reconstruction, <Miriad-for-SMA> < d-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miria -SMA> <Miriad-for-SMA> Image <Miriad-for-SMA> Analysis and Display <Miriad-for-SMA> <Miriad-for > <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA iriad-for-sma> <Miriad-for-SMA> Users <Miriad-for-SMA> Guide <Miriad-for-SMA> <Mi for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-f A> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SM Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> < d-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miria -SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for > <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA iriad-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Mi for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-f A> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SM Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> < d-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miria -SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for > <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA iriad-for-sma> <Miriad-for-SMA> Jun-Hui <Miriad-for-SMA> Zhao <Miriad-for-SMA> <Mi for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-f A> <Miriad-for-SMA> <Miriad-for-SMA> 09 Jul 2012<Miriad-for-SMA> <Miriad-for-SM Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> < d-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miria -SMA> <Miriad-for-SMA> See <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for > <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA iriad-for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Mi for-sma> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-for-SMA> <Miriad-f

3 Contents Table of Contents 1 MIRIAD Software for SMA Introduction Acknowledgements Management and Support Rules in MIRIAD Programming List of SMA Codes DATA FORMAT TRANSFER SMA SPECIFIC CORRECTION DATA INSPECTION AND EDITING WITH X-WINDOW GRAPHICS CALIBRATION UTILITY SMA MIRIAD Beta Release Site & Installation MIRIAD ON SAO PUBLIC COMPUTERS Data Format from the SMA Correlator ENDIAN FORMAT FREQUENCY CONFIGURATIONS POLARIZATION Run MIRIAD INP GO EXIT QUIT UNSET SAVE ii

4 CONTENTS iii LOAD TPUT TGET MORE ABOUT MIRIAD COMMANDS Documentation in MIRIAD HELP DOC UV Data Loading, Display, Flagging and Corrections Introduction Demonstration LOADING DATA SCANS UV FILE CHECK AND ANALYSE UV VARIABLES SPECIFIC CORRECTION INSPECTION OF SPECTRAL DATA PLOTTING VISIBILITIES FLAGGING DATA SMASPLT for 4 GHz Mode Utility Functions LINETYPE SELECTION POLARIZATION/STOKES Utility Tools INSPECTION OF VARIABLES BASELINE FITTING LINEAR REGRESSION BETWEEN 230/690 GHz PHASES PHASE RESIDUAL VS. AMBIENT TEMPERATURE GLUE SMA MULTI-CHUNK FILES TOOL FOR OBSERVING SCHEDULE SORTING UVDATA INTO A TIME ORDER Strategies in Calibrations Introduction

5 iv CONTENTS 3.2 SMA Spectral Data and Calibration Strategy BANDPASS AND GAIN SOLVERS Normal Calibration Procedure BANDPASS ANTENNA GAINS APPLY BANDPASS AND ANTENNA GAINS Special Calibration Tools and Procedures BANDPASS RIPPLES BANDPASS CORRECTIONS ACROSS SIDEBANDS BASELINE-BASED BANDPASS CALIBRATION WITH MULTIPLE CALIBRATORS PHASE TRANSFER BETWEEN 230 AND 690 GHz Calibration for The Flux Density Scale Hybrid Spectral Resolution Very High Spectral Resolution Export Calibrated Data FITS to AIPS FITS to GILDAS Basics in Imaging, Deconvolution and Self-calibration Introduction Weighting Dirty Image Deconvolution Self-calibration and Continuum Image Spectral Line Image Cube SPECTRA SPECTRAL IMAGE CUBE Synthesis Imaging with Multiple UV Datasets CATENATING SUBSETS OF UV DATA DATA ACQUIRED FROM MULTIPLE ARRAY CONFIGURATIONS DATA ACQUIRED WITH DIFFERENT CORRELATOR CONFIGURATIONS Application Utility (UV) 5-1

6 CONTENTS v 5.1 Introduction The JPL Catalog And Identification of Spectral Lines SCREEN INTERACTIVE SCRIPT AND MOLECULAR NAME TAGS Spectral Line and Continuum Emission Separation Amplitude Analysis SINUSOID BEATING OF TWO POINT SOURCES TIME VARIABILITY IN AMPLITUDE Re-Weighting Visibility Application Utility (IM) Introduction Primary Beam Correction for Single Field Primary Beam Correction and Mosaicing Multiple Fields Conversions between Coordinate Systems Scripts for SMA Data Reduction Introduction EXAMPLE: imscript.csh End-to-End Data Reduction MFS CONTINUUM IMAGE: contmscript.csh MAKE A SPECTRAL LINE CUBE: maserscript.csh TEST THE BASIC MIRIAD TASKS FOR SMA: cepha lsb cntm.csh Data Reduction Scripts for the Basic SMA Observing Modes A SMA UV data and Variables A-1 A.1 SMA Data in MIRIAD Format A-1 A.2 Variables A-1 B REFERENCES B-1 C A List of Miriad Tasks C-1 C.1 By Alphabet C-1 C.2 By Category C-3

7 vi CONTENTS

8 Chapter 1 MIRIAD Software for SMA 1.1 Introduction MIRIAD is a radio interferometry data reduction package. MIRIAD can be configured for the specifications required by the SMA. A few specific codes for the SMA have been developed and are supported. MIRIAD now can be used for the reduction of continuum and spectral line observations from loading of the original SMA archival data through to the image analysis. The SMA features have particular emphasis on aspects of interest to users of the SMA. We supports the off-line software for SMA data reduction under the MIRIAD evironment. The MIRIAD software components include tools used for calibration and analysis of the SMA data observed at submillimeter wavelengths. We have successfully implemented MIRIAD software for handling SMA data. Coping with the SMA correlator upgrade, SMA MIRIAD Beta Release Site has been constructed to frequently update MIRIAD software components for SMA data reduction. 1.2 Acknowledgements We thanks MIRIAD developers for supporting our users for the reduction of SMA data in MIRIAD. 1.3 Management and Support New software components developed in MIRIAD for the reduction of SMA data are considered as a part of the MIRIAD software which is managed by the Miriad group via CVS (Concurrent Versions System). Integration of the individual MIRIAD codes into the MIRIAD CVS is supported. The source codes for SMA specifications are placed under $MIRSRC/prog/sma. Special requirements and questions from SMA users can be sent to smamiriad@cfa.harvard.edu. The update components in the SMA beta version will be integrated back to MIRIAD CVS. 1.4 Rules in MIRIAD Programming SMA users are encouraged to make their own routines for special needs in the reduction of SMA data. The routines can be integrated into MIRIAD if the programming rules in MIRIAD are followed. For 1-1

9 1-2 CHAPTER 1. MIRIAD SOFTWARE FOR SMA example: The system variable $MIR is assumed to be defined as the path to the ordinary MIRIAD. One can do Pre-Compiling using ratty and checks source-code using flint: flint -I $MIRINC smaprog.for ratty -b -s -I $MIRINC smaprog.for smaprog.f ratty converts certain language extensions used in the original MIRIAD codes into standard FORTRAN. The command line flags are recognized as: -b If this flag is given, every backslash in the input is converted to two backslashes in the output. This is useful when the target compiler treats backslash as an escape character. -I The string following this flag indicates an alternate directory to search for include files. -s The string following this flag indicates the target compilers. Don t forget to modify the inline documentation which is usually placed in the beginning of each program and then make an inline doc file: doc smaprog.for After editing the code, one can export the source code smaprog.for to $MIRSRC/prog/sma and the documentation code smaprog.doc to $MIRDOC/prog. To compile the code and export the executable code, one can get an example of script and Makefile (smaprog.csh and make smaprog) from MIRIAD Test Codes for SMA Specifications and change the code name in the script to match your code. Then run the script. An example of codecompiling scripts is given in smaprog.csh. 1.5 List of SMA Codes In general, the MIRIAD codes can be configured for several telescopes. A terminal message beta... for handling SMA 4GHz and 2GHz mode data should be prompted in login, infoming that the beta SMA MIRIAD codes have been configured for the reduction of SMA data in the aspect of the configuration of the spectral data produced from the SMA digital correlator. The MIRIAD codes for the additional specifications required by the SMA users for calibration and data analysis are prefixed with sma. A list of MIRIAD routines with additional features for handling SMA data are given here DATA FORMAT TRANSFER smalod - converts the original SMA data format MIRIAD format; Keywords in smalod are described from the inline documentation.

10 1.5. LIST OF SMA CODES SMA SPECIFIC CORRECTION smafix - makes SMA specific corrections for quantities such as Tsys, and etc.. smasfreq - corrects for sky-frequency labelling error in the old SMA data (before ) smasplt - splitting the uv dataset to separate spectral chunk data from the first single channel chunk (ch0). Keywords in smafix, smasfreq and smasplt are described from the inline documentation DATA INSPECTION AND EDITING WITH X-WINDOW GRAPHICS smablflag - is an interactive flagger; plots a baseline as a function of time and flags bad points using the plotting cursor (color coded for source). smacheck - examines uv variables for SMA data and flags uv-data if variables or data are out of range. smauvplt - plots visibility data (color coded for sources) as a function of time and other variables (used for non-polarization data). smauvspec - plots SMA spectra (color coded for chunks); reads JPL spectral line catalog and plots molecular transitions along with SMA uv spectra. smavarplt - plots variables vs variables (color coded for sources). Descriptions of the keywords in the SMA routines ( smacheck, smauvspec, smauvplt, smavarplt, and smablflag) can be found from the inline documentation CALIBRATION phatrans - transfers phase correction from gains derived from a vis data observed at a lower frequency (freq1) to the phase corrections for the vis data observed simultaneously at a higher frequency (freq2). smaflux - sets the flux density scale of a visibility dataset

11 1-4 CHAPTER 1. MIRIAD SOFTWARE FOR SMA given a planet observation. smagpplt - plots and lists antenna gains, delay terms and bandpass; there are options of smoothing or polynomial fitting to the calibration curves. smamfcal - calculates calibration corrections (antenna gains, delay parameters and bandpass shapes) from a multifrequency observation; supports features for pre-processing uv data prior to bandpass/gain solving, such as continuum vector normalization, weighting, and smoothing. smatbpass - removes time-dependent, antenna-based bandpass ripples. Descriptions of the keywords in the SMA calibration routines ( smamfcal, smagpplt, phatrans, smatbpass and smaflux ) can be found from the inline documentation UTILITY schedobs smablsrc smacatool - plots UT time vs. EL to schedule observations of sources. - creates a uv dataset of pseudo-point sources, dividing the channel data by the continuum vector. - reads miriad calibrator catalogs and plots SMA and CARMA data together smachunkglue - glues together individual multi-chunk files into one mega-multi-chunk file. smauvamp smarewt uvsort - analyses visibiliy amplitude. - Copy a uv dataset, multiplying the visibility variances by the mean antenna gains for re-weighting the visibility. - copies a uv dataset, sorting uvdata into ascending order in time. Descriptions of the keywords in the SMA routines ( schedobs, smablsrc, smachunkglue, smauvamp, smarewt and uvsort) can be found from the inline documentation. 1.6 SMA MIRIAD Beta Release Site & Installation SMA Miriad Beta Release Site provides an updated Miriad software package for users to carry out SMA data reduction. Regular and binary installations for Redhat & CentOS are supported. Users are also encouraged

12 1.7. DATA FORMAT FROM THE SMA CORRELATOR 1-5 to install the software package on other operating systems MIRIAD ON SAO PUBLIC COMPUTERS The SMA MIRIAD beta version has been installed and updated on the public Linux computer systems at CfA, including Radio-Geoastronomy division, SMA Hawaii site and Computation Facility. 1.7 Data Format from the SMA Correlator The SMA Data File Format has been used to format and store the visibility data along with the ancillary data in a set of files all contained in a single directory with a name identified with the start time of an observing track, e.g : 51 : ENDIAN FORMAT The online data has been processed with a big endian computer PowerPCs. The online processor will be soon replaced with a little endian computer. The conversion of the online endian format to that of users desktop/laptop computers are taken care by the task smalod in MIRIAD FREQUENCY CONFIGURATIONS In addition to the visibility data, the online processor stores the the header information for each individual spectral band of data into a file sp read under the directory of a data file, containing a set of spectral bands which can be either a correlator chunk (spectral window) or the pseudo-continumm channel ( a single channel that is a vector-average from all the channels of the spectral bands). In history (prior to 2009 spring), the SMA frequency configuration consists of one continuum channel (c1) and 24 spectral bands (s1 to s24) with a total bandwidth of 2 GHz for each sideband; this correlation operation mode is often referred as a 2 GHz mode. For a uniform resolution mode, all the 24 spectral chunks have the same channel resolution. For a hybrid spectral resolution mode, some of spectral chunks may be empty. Note that after 2009 spring, this 2 GHz mode has been occasionally used for high spectral resolution observations. The SMA Software Limits have been followed to program the data reduction software used for the old standard. Since 2009 spring, the IF bands have been doubled. The double-band frequency configuration then consists of one continuum channel (c1) and 48 spectral bands (s1 to s48) with a total bandwidth of 4 GHz for each sideband. This double-band mode is also referred as a 4 GHz mode. In the near future, the IF bands will be doubled again with a total bandwidth of 8 GHz. In addition to previous 48 spectral bands, SMA will add additional two broad (wide) spectral bands with a total of channels and 2 GHz bandwidth for each. Thus, the new SMA frequency configuration will consists of one continuum channel (c1), 48 spectral bands (s1 to s48) and two wide bands (w1 and w2). Therefore, each sideband will have a total of 51 spectral bands. New SMA Software Limits have been discussed recently. The new standard will be followed in MIRIAD programming for SMA data reduction POLARIZATION MIRIAD supports polarization observations with the SMA. So far, circular-polarization observations using single receivers with quarter-wave plates sample LCP and RCP in a Walsh cycle for the SMA correlator to

13 1-6 CHAPTER 1. MIRIAD SOFTWARE FOR SMA produce four polarized correlations (LL, LR, RL, and RR) on each baseline. The reduction of the polarization data produced from this mode has been supported in MIRIAD. 1.8 Run MIRIAD After installation of the MIRIAD package, one may need to set the environment. At the beta site for SMA Miriad, installation notes are provided. After setting up the environment, one can enter MIRIAD by typing: jupiter.cfa.harvard.edu% miriad Then a prompt miriad% pops up indicating that one has entered MIRIAD environment. The MIRIAD shell provides a user interface. There are a few useful MIRIAD commands in the MIRIAD front-end: INP The MIRIAD command inp is to inspect the inputs of a task, as well as to select the task, e.g. smalod, type miriad% inp smalod MIRIAD will list the input parameters of the task, for exmaple: Task: smalod in = out = rxif = restfreq = vsource = refant = options = rsnchan = spskip = sideband = nfiles = nscans = GO The command go is to run a MIRIAD task. Tasks are run either by typing go taskname at the miriad% prompt (advisable only if you know you like the inputs) or by typing go at the taskname% prompt: smalod% go EXIT The command exit exits MIRIAD and, if any parameter values have been changed from the previous time you ran MIRIAD, all parameter values are saved in a keyword file lastexit. The parameter values

14 1.8. RUN MIRIAD 1-7 from this file are read in automatically when you next run MIRIAD QUIT The command quit exits MIRIAD without saving the present parameter values in lastexit. This command is useful if you have changed parameter values you do not wish to save UNSET It is often convenient to assign the original default value to the parameter. This assignment is accomplished with the command unset. For example, if the task smalod had been run with inputs: Task: smalod in = /home/miriad/smadata/sma/040329_09:37:37/ out = mymiriad.uv rxif = 1 rsnchan = 16 sideband = 0 nscans = 35,950 but one wanted to run smalod with the default for rsnchan, you would type: histo% unset rsnchan SAVE When one exits from MIRIAD all the inputs will be saved in the keyword file lastexit. However, one can save and reload the current parameters at any time. The command save writes the current parameter values to a keyword file (taskname.def if the task taskname has been selected). Note that save writes out all parameter values, not just those for the specific task chosen. Example: smalod% save smalod% save smalod1.pars In the first case the parameters are saved in a file smalod.def, in the second case they are saved in a userspecified file smalod1.pars LOAD The parameter settings save in the keyword file can be retrieved using the command load TPUT When one invokes a task, MIRIAD saves all the parameters associated with this task in a keyword file, in the directory given by the MIRDEF environment variable (if MIRDEF is not set, the current directory will be used). The parameters will be saved in a file with name taskname.def. Additionally you can force parameters to be saved using tput taskname. If the taskname is omitted, the parameters for the current task are saved.

15 1-8 CHAPTER 1. MIRIAD SOFTWARE FOR SMA smalod% tput smalod TGET The tget command is used to retrieve the parameters at a later stage. For example, to restore parameters, use smalod% tget smalod tget normally expects the keyword files to be in the directory given by the MIRDEF environment variable MORE ABOUT MIRIAD COMMANDS The Chapter on The User Interface in MIRIAD Users Guide (ATNF) is recommended for further reference on MIRIAD commands. 1.9 Documentation in MIRIAD Documentation is built into the MIRIAD source code, and extracted using the help command or the doc program HELP The help command is generally used to get information about a task. The general format of the command is: miriad% help [-w] taskname [-k keyword] For example: miriad% help smalod will give you help information about the MIRIAD task, smalod DOC The task doc extracts on-line documentation and presents the output in a variety of formats: miriad%doc telepar The output from doc telepar is: Task: telepar Responsible: Bob Sault

16 1.9. DOCUMENTATION IN MIRIAD 1-9 TELEPAR gives the characteristics of various observatories. Its main use is to check that the characteristics are correct. Keyword: telescop Name of the observatory. Several can be given. If none are given, TELEPAR simply lists the known observatories.

17 1-10 CHAPTER 1. MIRIAD SOFTWARE FOR SMA

18 Chapter 2 UV Data Loading, Display, Flagging and Corrections 2.1 Introduction The SMA online data computer takes spectral data from correlator and store the multiple spectral-band data into a file following the SMA Data File Format. The frequency configurations of SMA data are complicated and flexible. The SMA Passband Visualizer provides web-based graphic tool to set up frequency configuration across the SMA passband with a total number of spectral bands per sideband 25, 49 and 51 for the 2 GHz, 4GHz and 8 GHz (near future) modes, respectively. A spectral band here can be either a correlator chunk (spectral window) or the pseudo-continumm channel ( a single channel that is a vector-average from all the channels of the spectral bands). In principle, the number of channel and channel width (or spectral resolution) in each of the spectral chunks can be different. The MIRIAD along with SMA software features supports such flexibility in the configuration of spectral data. In order to handle a large number of spectral windows in MIRIAD a few of original MIRIAD codes are modified by adding new software features. To distinguish the spectral data between different spectral chunks (or windows) and different sources, both spectral windows and observing sources are coded with colors in various displaying routines. The following sections show demonstrations of processing SMA data in MIRIAD. 2.2 Demonstration A demonstration of routines for the handling of SMA data in MIRIAD is given here. One can run a demonstration by following the input parameters for each task and taking the raw data from the SMA archives. This demonstration will give users a taste on how the software routines are operated in MIRIAD and what functions are contained in these routines. Certainly, users are encouraged to explore more functions that are supported and to require for other specific features that may not have been supported. In particular, users can ask to include in the MIRIAD-formatted SMA data more variables that are needed for the reduction. The following demonstration shows the basic procedures and setups in processing SMA uv data. 2-1

19 2-2 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS LOADING DATA The task smalod converts the uv data file from the SMA archival format into MIRIAD format. Considering the complication and flexible of the SMA correlator configuration, smalod loads all the basic modes supported by the SMA. A few exmaples of usages are given below. Loading Non-polarization Data For observations with single receiver system, the original SMA data file will be loaded into MIRIAD in two separated files: gc rx1.lsb and gc rx1.usb for the lower and upper side bands, respectively. In calibration, the lsb and usb data sets are treated separately. For continuum imaging, both calibrated lsb and usb can be merged in the imaging process. smalod% inp Task: smalod in = /sma/data/040329_09:37:37/ % import raw data in SMA archival format. out = gc % prefix of output file name in MIRIAD format. rxif = 1 % rx/if id =1 sideband = 2 % lower side band. nscans = 35,950 % skip the first 35 integration scans. % and load the next 950 scans. smalod% inp Task: smalod in = /sma/data/040329_09:37:37/ % import raw data in SMA archival format. out = gc % prefix of output file name in MIRIAD format. rxif = 1 % rx/if id =1 sideband = 0 % lower side band. nscans = 35,950 % skip the first 35 integration scans. % and load the next 950 scans. For observations with dual receiver system, selection above will load only the lower side band data of one receiver (with a receiver id=1) from the original SMA data file. A total of 4 seperated files will be splitted from a raw SMA data file observed with a dual receiver system. MIRIAD supports calibration of the splitted data based on each of frequency bands. Since the version Oct-05 or a later version, smalod has been implemented with a function of reading the online flagging information by default. In addition, in case of the corrupted baseline vectors in the SMA data header which is used to decode the antenna positions of the array, there is an input Keyword: readant which allows users to read the ASCII file antennas in the SMA file data directory if readant=8 or larger number is specified. Here is an exmaple of the usage: smalod in=050911_04:21:35 out=050911h30a rxif=0 \ sideband=1 nscans=5, readant=8

20 2.2. DEMONSTRATION 2-3 Note that for the data taken from 4 GHz mode, smalod reads the antenna positions directly from the ASCII file antennas. Loading Circular-polarization Data For the polarization SMA data taken from single receiveris with quarter wave plates, an (options=circular) in smalod has been supported to load circular polarization data into MIRIAD: smalod in=050609_04:34:18/ out= rxif=0 refant=6 \ options=circular sideband=1 nscans=3 Loading Old Data The old data (say 2003) might have been taken from the incompleted SMA correlator. MIRIAD smalod with spskip=-2003 now supports loading old data produced from the partially completed correlator: smalod in=030910_03:53:34 out= rxif=1 \ sideband=1 nscans=44, spskip=-2003 Note that this feature is invalid for the data taken from 4 GHz mode. Swap Endian The conversion of endian format between SMA archival data to user s computers is taken care at the system level. MIRIAD task smalod set the default for big endian SMA archival data. For little endian SMA archival data, smalod provides an option (options=sendian) for converting little endian format to users computer: smalod in=120611_19:01:41 out=lendian.uv rxif=0 \ options=sendian sideband=1 Note that smalod prepares to handle little endian SMA data as default as soon as the SMA online data computer is replaced by a little endian machine. Flagging Ipointing Recently, SMA online storage software implements a feature for flagging ipointing data with non-zero offsets from the calibrators. This feature has been incorporated in smalod as a default SCANS UV FILE Task: uvindex vis = gc_rx1.lsb The task uvindex scans a uvdata file and reports changes in source name, pointing center, number of wideband and spectral line channels, the observing frequency, and polarization. This is a useful program and is recommended for users to run it immediately after the SMA data loaded in MIRIAD. It is necessary to check-up the basic information about the observation before further processing the data. In addition, the task prthd gives the basic information in the header of a uv data file.

21 2-4 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS CHECK AND ANALYSE UV VARIABLES MIRIAD task smacheck can be used to check uv variables for SMA data and flag uv-data if the corresponding variables or data are out of range. Here is an exmaple to check the variable systemp, give a report on simple statistical analysis, and flag the uv data if systemp is outside the range as specified. Task: smacheck vis = gc_rx1.lsb flagval = flag var = systemp range = 1,1200 options = histo In the case of polarization observations, SMA archival data might also contain non-polarization data (ipol=0), in addition to the four polarization components. The MIRIAD smalod loads every data point from SMA archival data by converting MIR polarization state label (ipol) to the MIRIAD convention and retaining the non-polarization data unchanged. The non-polarization data might cause problem to the polarization calibration process in MIRIAD and needs to be flagged. For the circular polarization case, smacheck can flag the non-polarization data: Task: smacheck vis = uvtest_rx1.lsb flagval = flag options = circular SPECIFIC CORRECTION Amplitude Correction and System Temperature A major error in the data observed at submillimter wavelengths is the phase distortion and amplitude attenuation due to the airmass in the atmosphere. The system temperature (T sys ) measurements determined by comparing the receiver output power from an ambient temperature load with that of the sky are essential to tracking amplitude attenuation variations. Therefore, the system temperature (T sys ) can be used to correct for the gains in amplitude. No system temperature corrections have been applied to the SMA archival data. However the frequently measured T sys values for each antenna are stored in the header of SMA archival data for each pair of antennas under the baseline file bl read. The antenna-based T sys are computed and stored in the header of uvdata in MIRIAD by smalod. The measured T sys is used to calibrate the atmosphere attenuation in the visibility amplitudes. Correction for System Temperature Corrections for system temperature are carried out offline. For the recent SMA data, the corrections for T sys use smafix : smafix vis=050911h30a_rx0.usb out=050911h30a_rx0.usb.tsys \ device=/xs xaxis=time yaxis=systemp nxy=2,4 \ options=tsyscorr

22 2.2. DEMONSTRATION 2-5 Here, H30a rx0.usb is the MIRIAD formated data that is directly converted from SMA archival data and H30a rx0.usb.tsys is the output file which has been applied for T sys to the fringe amplitude of the visibility. For the older SMA data, since bad stablity of the total power detectors in T sys measurements, the values of system temperature were corrupted. a repair for the corrupted T sys measurements might be required prior to applying the T sys corrections to the visibility data. smafix provides features for fixing bad T sys measurements. For old SMA archival data, here is a recommended procedure for T sys correction in MIRIAD : 1. Polynomial Fit to T sys and Apply T sys Correction smafix% inp Task: smafix vis = gc_rx1.lsb % name of the visibility data file; out = gc_rx1.lsb.tsys % name of output file; device = /xs % pgplot device,x-windows here; xaxis = time % time in xaxis; yaxis = systemp % system temperature in yaxis; nxy = 2,4 % 2 plots in xaxis and yrange = 100, in yaxis; % plot range from 100 to 1500 in yaxis; dofit = 2 % polynomial fit; options = tsyscorr,dosour,tsysswap % apply tsys corrections, do source based fitting, replace the raw Tsys with fitted values. A few fitting options have been provided to derive the T sys corrections. Fig.2.1 shows a plot of antenna-based system temperature. Each panel stands for one antenna. A second order source-based polynomial fit to T sys (solid lines) are carried out in order to get rid of the bad T sys measurements. If options=tsyscorr, then the polynomial curves are replaced in the correctionis for system temperature. Careful users might like to check up the polynomial curves. The fitted T sys is stored as a variable systmp in the uv data in the defualt options. If options=tsysswap, the fitted T sys will swap with the orignal T sys, i.e. the fitted T sys is stored as a variable systemp and the original T sys goes to variable systmp. 2. Check and Plot Fitted T sys One can plot systmp or systemp using varplt or smavarplt (the sources are coded in different color; see Fig. 2.2): smavarplt% inp Task: smavarplt vis = gc_rx1.lsb.tsys device = /xs xaxis = time yaxis = systemp % the variable for fitted Tsys nxy = 2,4

23 2-6 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.1: This plot shows a source-based second order polynomial fit to the antenna-based T sys as function of time for 8 antennas.

24 2.2. DEMONSTRATION 2-7 Figure 2.2: The ploynomial fit to system temperature is plotted. Each panel stands for each antenna. It is worth to mention that other MIRIAD tasks such as invert, smamfcal will use the variable values of systemp to calculate the variance of the data and the sensitivity weight in imaging and calibration.

25 2-8 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS 3. Replacement and Flagging of Corrupted T sys Occationally, an antenna might be corrupted in T sys measurements. Users can choose a good antenna in T sys measurements from the rest of the antennas to replace the bad antenna. The keyword bant in smafix can specify the id of the corrupted antenna and gant assigns a good antenna to replace the bad one. Here is a usage: Task: smafix vis = gc_rx1.lsb out = gc_rx1.lsb.tsys device = /xs xaxis = antel yaxis = systemp nxy = 2,4 bant = 4 % ant 4 is bad. gant = 1 % Tsys values of the good ant 1 will replace those of ant 4. dofit = 2 options = tsyscorr,dosour,tsysswap Alternatively, the systemp values of bad antennas can be replaced with those of a good antenna without performing polynomial fitting: Task: smafix vis = visdata out = visdata.tsys device = /xs xaxis = time yaxis = systemp nxy = 2,4 bant = 1,8 % ants 1, 8 are bad. gant = 3 % Tsys values of the good ant 3 will replace those of ants 1, 8. If only a few data points are corrupted, one may use smacheck to flag the data outside a specified range for the T sys or other variables. By default, smafix will only take the unflagged (good) data in fitting, plotting and correction. There is an options of all to allow to use all (flagged and unflagged) data in the T sys correction. Correction for Sky Frequencies In all the SMA data observed before , there is a labelling error for the sky frequencies that has lead to 1/2 of the chunks labelled with frequencies 1/2 channel step too high, and the other half labelled with frequencies 1/2 channel step too low. With the updated version of smalod (version Nov-07 or later), the labelling problem has been automatically taken care. For the SMA data fully calibrated in MIRIAD, a task smasfreq can be used to fix the labelling error if the data in MIRIAD format keep the information of the original frequency configuration, i.e. 24 spectral windows per sideband with the original channel resolutions. Here is an example to use smasfreq: %apply gain/bandpass solutions to the visibilities if the corrections have not applied yet: uvaver vis=071123_20_rx0.lsb out=071123_20_rx0.lsb.calibrated

26 2.2. DEMONSTRATION 2-9 %then correct for the sky frequency errors: smasfreq vis=071123_20_rx0.lsb.calibrated out=071123_20_rx0.lsb.calibrated.sfreq The task will prompt messages on screen while the program is excuted: SMAsfreq: Version 28-Nov-07 Processing input file _20_rx0.lsb.calibrated Updating the sky frequencies: chunkid nchan in-sfreq(ghz) out-sfreq(ghz) sdf(mhz) s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => s => Copy over 390 records from _20_rx0.lsb.calibrated Correction for Velocity Information In the case of a Doppler tracked observation, the sky frequency slowly changes with time due to the motion of the observatory with respect to the rest frame chosen. Various velocity components need to be taken into account for a particular rest frame. For example, in the Local Standard of Rest (LSR) frame, three components are included in the Doppler tracking: Components Correct for Speed Diurnal Observatory motion due to < ~0.5 km/s the Earth s rotation Annual Earth s motion around the < ~30 km/s Sun Solar motion Solar motion w.r.t. the LSR ~20 km/s

27 2-10 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS In order to reduce the observed velocity with respect to the rest frame, one needs the information of the sky frequency and the Doppler velocity projected in the direction of the source in addition to the rest frequency. At the SMA, two observing modes have been supported: 1) non-planet mode and 2) planet mode. For the non-planet observing mode, the SMA supports on-line Doppler tracking for spectral line observations, i.e. the observatory continuously changes the observing frequency to account for the effect of the observatory motion with respect to the rest frame. However, a sky frequency with a fixed value at the central channel of a given spectral chunk is stored in the archived SMA data. The sky frequency in the SMA data has been corrected for a part of the tracked Doppler velocity (the diurnal term and a part of annual term). Therefore, the sky frequencies given in the SMA data appear not to follow the convention of MIRIAD. For the BIMA array, on-line Doppler tracking was supported and the Doppler tracked frequency was stored in the data header, slowly varying with time (MIRIAD Users Guide for BIMA, version 1.0 March 1991, printed: August 3, 2004; and inline note in uvindex.for by Mel Wright). For the ATCA, the observatory does not Doppler track. The channel-smearing problem due to the Earth motion is handled in MIRIAD by converting the frequency channels to velocity channel (MIRIAD Users Guide for ATCA by Bob Sault and Neil Killeen, 19 Mar 2004). Regardless of whether the observatories Doppler track or not, a particular velocity component will remain in the same velocity channel using the linetype in MIRIAD : line=velocity,nchan,start,width,step The usage of the MIRIAD linetype requires that the sky frequency and the Doppler velocity provided in the data are self-consistent. In order to use the MIRIAD software for the reduction of SMA spectral line data, SMA users might consider to recompute velocity information depending on the following three cases. The corrections are particularly important to the observations of narrow spectral lines. 1. Deriving Velocity Information with smalod In the case that the Doppler tracked source is the only interested target source (e.g. SgrB2 N) in one s program (e.g :49:59) and the sky frequency stored in the archived SMA data has been corrected for a part of the Doppler velocity, one may use smalod in the defualt options to derive the proper residual Doppler velocity ( V D ) for the target source (SgrB2 N): V D = V D V SMA, where V D is the tracked Doppler velocity and V SMA is the effective part of the tracked Doppler velocity that has been taken out in the on-line calculation of the sky frequency which is stored in the SMA data. We note that in MIRIAD the Doppler velocity V D (veldop) is the sum of the radial velocity of the observatory (in the direction of the source, with respect to the rest frame) and the nominal system radial velocity of the source (vsource). MIRIAD task smalod decodes the radial velocity of the observatory from the information in the SMA archival data header files. The value of vsource is given by users from input Keywords. As long as the value of the uv variable vsource is consistent with veldop in the MIRIAD data header, the MIRIAD linetype and tasks will correctly handle the conversion between radial velocity and sky frequency. Here is the usage of smalod for this case: Task: smalod in = /data/smadata/050801_04:49:59/ % name of input file out = smatest % prefix of the output file rxif = 0 % rx id vsource = % the nominal velocity of target source w.r.t. the rest frame

28 2.2. DEMONSTRATION 2-11 Figure 2.3: The value of the Doppler velocity ( V D ) derived from the SMA archival data header corresponding to the stored sky frequency as function of UT time is plotted. This value ( V D ) is correct only for the Doppler tracked target source (SgrB2 N). refant = 6 sideband = 0 nscans = 2, which was applied in the online Doppler tracking. Here we used default value to set vsource to zero. % reference antenna % lower side band % skip the first two integrations The derived value of the residual Doppler velocity V D corresponding to the stored sky frequency (sfreq in MIRIAD ) is saved in the MIRIAD variable veldop in the header of the output MIRIAD uv data (smatest rx0.lsb for this case). To inspect the derived residual Doppler velocity V D, one can use smavarplt to plot the variable veldop as function of ut time: smavarplt vis=smatest_rx0.lsb device=/xs xaxis=time yaxis=veldop nxy=1,1 One should see a constant value across the observing period for all the sources. Keeping in mind, this value is correct only for the Doppler tracked target source (SgrB2 N). 2. Recomputing Velocity Information with uvredo Besides the Doppler tracked target source (e.g. SgrB2 N), other sources (e.g. SgrB2 M and sgrastar) might be interested in the observing program. One might need to recompute the velocity information in order to avoid errors in reducing the radial velocities of the sources and to get rid of channel-smearing effect in the synthesized spectral line images due to the errors in velocity. A patch for handling SMA data has been added into uvredo. The SMA patch calculates the proper residual Doppler velocity ( V D ) for each of the data records corresponding to the sky frequency in the data. Here

29 2-12 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.4: The proper residual Doppler velocities ( V D ) recomputed with the SMA patch in uvredo are plotted for all the sources. is the usage of uvredo for using the SMA patch: Task: uvredo vis = smatest_rx0.lsb % input data out = smatest_rx0.lsbrd % output data options = smaveldop % using the SMA patch to recompute the velocity information dopsour = SgrB2_N % the name of the Doppler tracked source doptime = % default using smavel to calculate the part of Doppler velocity that has been corrected to the sky frequency velocity = lsr % rest frame The results on the velocity information recomputed using the SMA patch in uvredo can be reviewed with a plot or a log file using smavarplt: smavarplt vis=smatest_rx0.lsbrd device=/xs log=log xaxis=time yaxis=veldop nxy=1,1 Fig. 2.4 shows the proper residual Doppler velocities for the sources corresponding to the sky frequency in the data. The numerical values at each of the records plotted can be reviewed from the log file (log). We note that the radial velocity of a planet with respect to the rest frame might significantly change during the course of a long observing track. For the use of non-planet observing mode, planets are used as calibrators with a short observing time on source and the effect due to the motions of planets is ignored.

30 2.2. DEMONSTRATION 2-13 After having the proper residual Doppler velocities corresponding to the sky frequency in the data header, we can safely use the MIRIAD tasks to reduce the SMA spectral line data. 3. Case without On-line Correction to The Sky Frequency In the case that Doppler tracking is performed but no on-line corrections are made to the sky frequency, the frequency stored in the data header is the true sky frequency. The frequency in the SMA data is considered to follow the MIRIAD convention. Then, one needs to simply recalculate the Doppler velocities while loading the SMA data to MIRIAD using smalod. smalod in=/data/smadata/050101_03:59:50 out=smatest \ rxif=1 vsource=-30 refant=6 options=lsr sideband=1 nscans=2, With options=lsr, the Doppler velocities can be calculated for each of the sources and are stored in the MIRIAD variable veldop. Alternatively, one can recalculate the Doppler velocity with options=velocity in uvredo: uvredo vis=smatest_rx1.usb out=smatest_rx1.usbrd options=velocity velocity=lsr 4. Reset The Value for The Radial Velocity (vsource) We note that uvredo recomputes only the radial velocity of the observatory in the direction of the source, with respect to the rest frame and resets vsource to be zero, i.e. the nominal radial velocity of the source is automatically taken out from both veldop and vsource. In the case of using smalod, the Doppler velocity veldop decoded from the MIR data header includes only the radial velocity of the observatory in the direction of the source, with respect to the rest frame. Users might want to add the systemic radial velocity of the source back. Then one must give the value of the radial velocity of the source to vsource in the input Keyword of smalod. Alternatively, with puthd, users can put a value for vsource puthd in=psource_rx1.usb/vsource value=-11 type=real The radial velocity of 11 km/s is put to vsource in the header of the data. Users must beware of using puthd to reset vsource unless ones know what is the consequence caused by the change of vsource. We recommend that users run uvredo to avoid the complication in getting the correct velocity information. Correction for Rest Frequency In the case of multiple spectral line observations, users might want to change the rest frequency corresponding to the interested line transition. puthd in=miriad.uv_rx1.usb/restfreq value= type=double Correction for uvw coordinates Antenna positions might be not accurately measured while the array is reconfigured. The uvw coordinates calculated in on-line software might be subject to significant errors which will degrade the image quality and must be corrected for.

31 2-14 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS The uvw coordinates can be fixed after the position offsets are determined or the updated antenna position file is provided. In MIRIAD the task uvedit now supports SMA users to recalculate the uvw coordinates. The usages of uvedit are provided for the SMA users in the following two cases: 1. An original antenna file and antenna offsets From the SMA log, users might be able to find the offsets from baseline-fitting program which gives the antenna position offsets usually in unit of the observing wavelength. The offsets need to be converted to in unit of millimeter to use uvedit. The original antenna file (antennas) porvided in the raw SMA data file need to be copied over. Here is an example of SMA antenna file ( :31:25): e e e e e e e e e e e e e e e e e e e e e e e e+00 The first column is the antenna identification number. The next three columns are the geocentric coordinates (xyz) in meter. Task: uvedit vis = uvtest_rx0.lsb %input vis file. source = 3c279 %the name of the source to apply the corrections to. apfile = antennas %the original antenna file. smaoffset= 1,0,1.5,0,4,2.2,0,0 %position offsets for the specific antennas; here are the case of: The antenna 1 has offsets of 0 mm in x, 1.5 mm in y, and 0 mm in z; the antenna 4 has offsets 2.2 mm in x, 0 mm in y and 0 mm in z. out = uvtst_rx0.lsb.uvw %output file with the corrected uvw. options = sma %options for sma specifications of geocentric coordinates system and the format of antenna file. 2. An updated antenna file If an updated antenna file with corrected antenna position coordinates is provided, the usage of uvedit is straightforward: uvedit vis=uvtest_rx0.lsb source=3c279 apfile=antennas \ out=uvcor_rx0.lsb.uvw options=sma

32 2.2. DEMONSTRATION INSPECTION OF SPECTRAL DATA The graphic display tool smauvspec was developed based on the original MIRIAD routine uvspec. The task, smauvspec, plots averaged spectra of a visibility dataset for all the SMA spectral chunks in one sideband with a color coded for each spectral chunk data or for specific spectral windows using the function select. Averaging can be made in both time and frequency along the x-axis. smauvspec% inp Task: smauvspec vis = gc_rx1.lsb.tsys % input name of the uv data stokes = xx interval = 3 % 3 min in average axis = chan,ampl % plot amplitude vs channel; or (chan, phas) plot phase vs channel. device = /xs % pgplot device: x-window nxy = 7,4 % 7 by 4 panels on each page this plots a snapshot of all the spectral data per page for all 28 baselines. log = With smauvspec users can inspect the spectral data in various ways. The above example shows a way to quickly go through all the data and to see if there is any bad data that might contaminate the spectra. Fig.2.5 shows the 24 spectral chunks (coded color) of Jupiter on 28 baselines. It is obvious that the baseline 5-8 has a problem. This problem is present on all sources during the observation. It turns out that the problem was due to a bad correlator chip of the first block of spectral chunks (s1,s2,s3,s4) on this particular baseline. Note that the SMA spectral band id (a number preceded with letter s) is coded in the SMA archival data but is not used in MIRIAD. The phase plot (Fig.2.6) also shows the problem. Now, we take a particular antenna (4) to have a close look at its phase. The input parameters of smauvspec below demonstrate an example to look at the phase on the antenna (4) from snapshot spectra of Callisto. A phase delay is clearly seen on the baselines related to antenna 4 (see Fig.2.7). The MIRIAD functions select and line give a variety of options to select data. Please read Section Utility Functions in this Guide or the original MIRIAD User s guide (both BIMA and ATNF versions) for the details. smauvspec% inp Task: smauvspec vis = gc_rx1.lsb.tsys select = source(cal*),ant(4) stokes = xx interval = 30 axis = freq,phas device = /xs nxy = 2, PLOTTING VISIBILITIES The MIRIAD task uvplt can plot a variety of quantities from visibility data for each baseline, which is a useful application tool for users to identify bad data points. A feature of color-coded for sources has

33 2-16 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.5: This amplitude plot shows a snapshot of all the 24 spectral windows on all the 28 baselines on Jupiter. Note that an obvious problem due to bad correlator chips on baseline 5-8 can be easily identified.

34 2.2. DEMONSTRATION 2-17 Figure 2.6: This phase plot shows a snapshot of all the 24 spectral windows on all the 28 baselines on Jupiter. Note that an obvious problem due to a bad correlator chip on baseline 5-8 can be easily identified.

35 2-18 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.7: This phase plot shows the wrapped phase as a function of frequency due to a delay on the baselines related to antenna 4.

36 2.2. DEMONSTRATION 2-19 been added to the task smauvplt. For polarization data, color index used for polarization components in the original code uvplt conflicts with the color index for source in some cases. For non-polarization data, smauvplt has an advantage to identify bad data points in a multiple source file. Here is an example to plot the visibilities versus time by selecting the baselines related to antenna 4: Task: smauvplt vis = gc_rx1.lsb.tsys select = ant(4) axis = time,ampl yrange = -10,250 device = /xs nxy = 2,4 The beginning visibility points in each of scans appear to be bad (Fig.2.8) FLAGGING DATA With uvspec/smauvspec and uvplt/smauvplt, we have identified some problems in the SMA data. The bad data points must be removed prior to further data processing. There are a variety of ways to flag bad correlations. MIRIAD provides several useful tools for SMA users to flag bad data points. A few of the MIRIAD flagging/editing tools are introduced in the following. Uvflag: Flagging Selected Chunks and Baselines The MIRIAD task uvflag is used to change flags corresponding to the visibility data. Using the keywords select, line and edge, one can select a part of the uv-data that have been identified with problems. For all selected correlations, the flags are then set to the value specified with the keyword flagval. An example is provided below of flagging the bad correlations shown in the data inspection process: uvflag% inp Task: uvflag vis = gc_rx1.lsb.tsys select = window(1,2,3,4),ant(5)(8) line = edge = flagval = flag options = log = The setup above is to flag the four spectral chunks (s1-4) on the baseline 5-8. Uvflag: Flagging Selected Antennas Sometimes, the data related to one particular antenna may be corrupted during the observation. The select function in MIRIAD provides a way to flag the antenna using uvflag: uvflag% inp Task: uvflag vis = gc_rx1.lsb.tsys

37 2-20 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.8: This plot shows the amplitude of the visibility as a function of time on the baselines related to antenna 4. The scans on different sources can be separated by the color. The problematic visibilities, such as the low gain in the beginning of each scan, can be easily identified.

38 2.2. DEMONSTRATION 2-21 select = ant(2) line = edge = flagval = flag options = log = Uvflag: Flagging Selected Sources in A Time Range The following example shows that the visibility data of Sgr A* in the UT time range between 0:0 and 10:0 is flagged. uvflag% inp Task: uvflag vis = gc_rx1.lsb.tsys select = source(sgra*),time(0:0,10:0) line = edge = flagval = flag options = log = Uvflag: Flagging Edge and Central Channels In uvflag, the keyword edge can be used to flag the edge channels and central channels if they are considered bad. There are three parameters may be given n, m and c. Flags will be changed only for the first n and last m channels of each spectral window and for the central c channels. For exmaple, edge = 15, 15, 2 This will lead the first 15, the last 15 and central 2 channels to change their flag status for every spectral windows. Uvflag: Flagging A Bad Channel There may be a problematic spectral channel. One needs to indentify the channel number using smauvspec, say, channel= 1567 shows a bad data point. In uvflag, one can specify this bad channel using MIRIAD linetype: Task: uvflag vis = uvdata.tsys select = line = channel,1,1567,1,1 edge = flagval = flag options = log =

39 2-22 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.9: A visibility plot (time vs. amplitude) on baseline 1-3 for the five sources observed. The bad correlations in the beginning scans are obvious. Interactive Baseline-Based Flagging We now flag the bad integrations indentified in uvplt or smauvplt. The task blflag is a baseline-based interactive flagger. For non-polarization data, smablflag in which the sources are coded with colors can be used. This task plots the visibilities (e.g. time vs. amplitude) either a baseline at a time or all at once, and allows users to flag discrepant points using the plotting cursor. The task blflag / smablflag appears to be very useful in identifying various bad correlations which hide in the spectra. For example, the bad correlations due to poor telescope pointing in the beginning integration scans can be easily picked up in the visibility plot with color-coded data points. The setup of the input parameters is given below: smablflag% inp Task: smablflag vis = gc_rx1.lsb.tsys line = device = /xs stokes = xx select = axis = time,ampl options = The defaults in line and select lead to average all channels in all the spectral windows and plot the visibilities (time vs. amplitude) on each baseline. Fig.2.9. shows a visibility plot (time vs. amplitude) on baseline 1-3 for the five sources observed in the program. The bad correlations in the beginning integration during repointing the array from one source to another are obvious. For example, the relatively higher amplitude in SgrA* and relatively lower amplitude in SgrB2 are clearly bad correlations. Users can flag bad data by moving the plotting cursor to the problematic visibility points and clicking the left button of the computer mouse. After marking all the bad visibilities (the red points) on one baseline (see the bottom panel of Fig.2.9), users can move to the next baseline by clicking the right button of the mouse. Users can go through all the baselines to flag bad correlations. The task smablflag only changes the flag status for

40 2.2. DEMONSTRATION 2-23 Figure 2.10: The bad correlations in the beginning scans are flagged (red points). the bad correlations (red points; see Fig.2.10) and no data points have been removed from the uv file in the flagging process. Any mistakes in the flagging process can be corrected and flagged data can be recovered. Interactive System Temperature Flagging Variables antel and systemp have been added to the axis selection in smablflag so that system temperature as function of either time or antenna elevation (or other variables supported) can be selected to do the interactive flagging. This is a useful feature for the SMA users to meticulously flag the bad data points as indicated by the Tsys measurements which are occasionally present. smablflag displays the geometrical mean of the Tsys measurements from two antennas for each of the baselines. Large discrepant data points in Tsys are likely subject to a certain problem and it would be wise to flag out the corresponding data. The interactive flagging of system temperature is, in general, recommended to be carried out before the systemp correction using smafix. Usage: smablflag vis=h30a_rx0.usb device=/xs axis=time,systemp or smablflag vis=h30a_rx0.usb device=/xs axis=antel,systemp Flagging Integrations near Change to Observing Setup Change of observing setup may cause bad integrations. In the SMA observations, occasionally after the source change, bad correlations may be generated in the first integrations before the instrument is correctly settled. If this occurs, SMA users may use MIRIAD task qvack to flag the bad correlations, e.g. qvack% inp

41 2-24 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Task: qvack vis = gc_rx1.lsb.tsys % uv file select = interval = 0.5 % time interval for which the data are flagged after a setup change force = options = mode = source % a change in the source observed Flagging Based on Variable Range MIRIAD task smacheck flag uv-data if variables or data are out of range. Here is a usage to flag uv data based on a T sys range: Task: smacheck vis = gc_rx1.lsb flagval = flag var = systemp range = 1,1200 UV data outside T sys range K will be flagged. Task: smacheck vis = gc_rx1.lsb flagval = flag var = inttime range = 7,15 UV data outside integration time (inttime) range 7-15 sec will be flagged. This will be useful to get rid of the pointing scans with integration time of 5 sec SMASPLT for 4 GHz Mode For the data taken with 4 GHz mode, the pseudo-continuum channel (c1) is loaded by smalod to place at the first spectral band along with the spectral chunks (s1 to s48). The pseudo-continuum channels are useful for interferometer pointing and other purposes. However, since c1 only contains a single channel, the MIRIAD tasks with edge might run into problems in handling spectral data by dropping a few low-gain channels at the band edges. The MIRIAD task smasplt is suggested to run prior to further data reduction, which splits the spectral chunks from the pseudo-continuum band: Task: smasplt vis = _rx0.usb out = _rx0.usb4GHz options = ifsall Here rx0.usb is the output data from smalod containing c1 (the pseudo-continuum band), s1-s24 (the spectral chunks from the first IF bands) and s25-s48 (the spectral chunks from the second IF bands). options = ifsall splits the all spectral chunks (s1 to s48) into the output file rx0.usb4ghz. With different options, smasplt gives selected data into the output file.

42 2.3. UTILITY FUNCTIONS Utility Functions In MIRIAD there are a few very useful utility functions embedded in a variety of uv tasks that support on-the-fly selection and averaging of channels, spectral windows and polarization correlations. In both BIMA and ATNF s user s guide, a description of these functions has been given in great details. In this section, an emphasis of these keywords on the SMA data reduction is discussed below. These utility functions appear to be very useful for SMA users to handle spectral data from the powerful SMA correlator. With an instruction, these functions can be utilized for the reduction of SMA data. The contents of the following description of these parameters follow those described in the ATNF s User s guide. Interleaving with some comments on the SMA specifications, the duplication would be convenient for SMA users to use this manual LINETYPE MIRIAD was originally designed for a telescope which simultaneously measured both spectral and continuum data using separate correlators. For SMA use, there is no real distinction between wideband (continuum) and spectral data. The MIRIAD distinction is not relevant. All the SMA data is treated initially as spectral data. Users can produce a continuum data by averaging the spectral data after certain corrections and calibrations. In addition, as has been mentioned in previous section, smalod supports a feature to resample the spectral data from high spectral resolution to low spectral resolution and from hybrid spectral resolution to uniform one. This is a good feature for calibration of antenna gains. The SMA data in MIRIAD are formatted into a set of spectral line windows or chunks in the nomenclature of the SMA; each window consists of a number of channels separated by a fixed increment in sky frequencies. A task prthd prints a header of SMA data in MIRIAD format, which gives a glance of the SMA spectral data structure in MIRIAD uvindex also gives the secptral configuration in addition to other information about the uv file. It is quite common, when analyzing, plotting or mapping visibility data, that one will want to perform some averaging and selection of the desired channels, and one might wish to examine either spectral or the continuum data. The ability to select a range of spectral channels, to perform averaging, and to resample in velocity is provided by the line parameter - also called the linetype. If one s data contain multiple spectral windows, one should also refer to the next section SELECTION for more information on spectral window selection. The linetype parameter consists of a string followed by up to four numbers. Defaults will be used for any trailing part of the linetype specification that is missing. The string can be one of: channel This gives raw or averaged spectral channels. This is generally the default if spectral data is present. Since most SMA data is treated as spectral data, this will be the most commonly used linetype. wide This gives raw or averaged wideband (continuum) data. This is the default if only wideband data are present. This will be of no interest to SMA users since no wideband data are produced. velocity This gives spectral data, that have been resampled at equal increments in radio velocity (or equivalently frequencies). The resampling operation is a weighted average of spectral channels. felocity This is like velocity, but allows the velocity parameters to be given using the optical definition. Note, however, that the resampling operation is still in equal increments in frequency (or equivalently radio velocity). Because of the difference between the radio and optical velocity definitions, equal increments in radio velocity are not quite equal in optical velocity, and visa versa. The velocity increment that one gives is used as the optical velocity increment of the first channel. The accompanying four numbers are used to specify the range of input channels selected and averaged to

43 2-26 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS produce the output channels. Four numbers are: nchan, start, width, step For channel and wide linetypes, start, width and step are channel numbers (channels are numbered from 1 to N), whereas for velocity and felocity linetypes these values are in km s 1 (the velocity is relative to the rest frame usually LSR). These values are: nchan The number of output channels produced. Generally it defaults to the maximum number of channels that can be produced from the input data. A value of zero can also be used to give you the default. start For channel and wide linetypes, the start value is the first input channel to be selected. For velocity and felocity linetypes, start is the center velocity of the first output channel to be formed. The default value is 1 channel. width This value determines the width of the selected channels. For channel and wide linetypes, this gives the number of input channels to average together to produce a single output channel. For velocity and felocity linetypes, this gives the velocity width (in km s 1 ) of the output channels. The default value is again 1 channel. step This parameter gives the increment between channels. For channel and wide linetypes, this gives the increment between selected input channels. For velocity and felocity linetype this gives the velocity increment between the output channels. This defaults to the same value as width. For example: line=channel,10 selects 10 output channels, being input spectral channels 1 to 10. Similarly line=channel,10,2,1,2 selects 10 output channels, starting at input spectral channel 2, and skipping every second input channel. If you wanted to average together every pair of channels (rather than skipping it), you would use something like line=channel,10,8,2,2 For SMA users, the following choice might be useful to form chunk averaged data for 24 spectral chunks and each chunks have 32 channels, line=channel,24,3,26,32 The above combination will produce chunk averaged (the central 26 channels) data for 24 spectral windows. Finally a linetype of: line=velocity,10,1.5,1.0,3.0 would return 10 velocity channels with velocities centered at 1.5, 4.5, 7.5, etc. km s 1. Each channel would have a width of 1 km s 1.

44 2.3. UTILITY FUNCTIONS 2-27 When using velocity, felocity or channel linetypes on datasets with multiple spectral windows, window selection, as described in the following sections, may be useful. Some tasks require two linetypes, the first being the linetype of the data, and the second the linetype of a single reference channel (see e.g. invert). When specifying a reference linetype, you do not give the nchan (it is always 1) or step (it makes no sense for a single channel) SELECTION In some observations, users might wish to process only a subset of the possible visibility data. In MIRIAD the visibility data to be selected are usually specified by one parameter the select keyword. This parameter is constructed from select a number of subcommands, each subcommand selecting or rejecting visibility data which satisfies some condition. The subcommands, which can be abbreviated to uniqueness, are as follows: time(t1,t2) This selects visibilities observed between times t1 and t2. Times are given in UT. The second time (t2) is optional. If missing, it is assumed to be 24 hours after t1. The time is composed of a date and time-of-day portions, either of which can be omitted (but not both at the same time!). When the date is omitted, then the selection matches data, for the given time-of-day, regardless of the day. This is most useful for files containing only a single day s data. When the time-of-day is omitted, then 00:00 is assumed. The times t1 and t2 are given in the form: yymmmdd.fff or yymmmdd:hh:mm:ss.s Here yy is the year, mmm are the first three letters of the month s name, dd is the day of the month, fff is a fraction of a day, hh is the hour (24-hour clock), mm are the minutes, and ss.s are seconds and fractions of a second. The time-of-day portion can be abbreviated. For example, the seconds part can be omitted. Indeed (provided a date is given) the time-of-day portion can be totally omitted if desired. Note that only a two-digit year is given. These two digits give a year in the century For example: 05jan12:12:30 represents 12:30 on 12 January, 2005, whereas 98jun03.5 represents midday on 3 June, 1998, and finally 04apr01 is April Fools Day in the year To give some examples:

45 2-28 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS select=time(05jan05:10:50,05jan05:17:20) will select data observed on 5th January, 2005, between 10:50 UT to 17:20 UT. The form where no date is given: select=time(10:50,17:20) will select data observed on any day between 10:50 and 17:20. To give an example where only one time is given, select=time(05jan03) will select all data observed on 3rd January, This form is only useful for a file containing several days of data. antenna(a1,a2,...)(b1,b2...) This selects according to the antennas and baseline. Here a1,a2,... is a list of antennas to select. The second list, b1,b2,... is optional. If present, only baselines corresponding to the antennas pair (a1,b1), (a1,b2),..., (a2,b1),..., etc are selected. For example, to select all visibilities using antennas 1 or 3, use select=antenna(1,3) To select all visibilities using baselines with antennas 1 and 2, or 3 and 2, use select=antenna(1,3)(2) uvrange(uvmin,uvmax) This selects visibilities whose u v radius is in the range uvmin to uvmax. If only one value is given, then uvmin is taken as zero. The units of uvmin and uvmax are kilowavelengths. uvnrange(uvmin,uvmax) This is the same as the uvrange subcommand, except that the units are nanoseconds. visibility(n1,n2) This selects visibilities numbered n1 to n2 inclusive. increment(inc) This selects only every inc th visibility. ra(hh:mm:ss1,hh:mm:ss2) Select visibility data within a particular range of right ascension. Right ascension is given in hh:mm:ss format, or as decimal hours. dec(dd:mm:ss1,dd:mm:ss2) Select visibility data within a particular range of declination. Declination is given in dd:mm:ss format, or as decimal degrees. dra(p1,p2) For data files containing several pointing centers, this selects visibilities whose pointing center is offset, in RA, from the main pointing center, by between p1 and p2 arcseconds. ddec(p1,p2) For data files containing several pointing centers, this selects visibilities whose pointing center is offset, in DEC, from the main pointing center, by between p1 and p2 arcseconds. pointing(p1,p2) For data files containing rms pointing error data, this selects visibilities with the rms pointing error in the range p1 to p2 arcseconds. If only one number is given, p1 is taken as 0. The rms pointing error of a visibility is taken as the maximum of the rms azimuth and rms elevation pointing errors of the two antennas. source(srcnam1,srcnam2,...) For data files which contain multiple sources, this selects data according to source name. Several source names can be given, separated by commas. The source name can include an asterisk (), which acts as a wildcard.

46 2.3. UTILITY FUNCTIONS 2-29 frequency(loval,hival) This selects data according to the sky frequency of the first channel. The loval and hival values give the permissible range, in GHz, of frequencies to accept. If only a single value is given, then this is used as a mid-frequency, and frequencies within a tolerance of 1% are accepted. amplitude(x,y) Any correlation, where the amplitude is between x and y, is processed. If only one value is given, it is assumed to be x, and y is assumed to be infinity. window(w1,w2,...) This can be used when the dataset contains multiple spectral windows (or chunks in SMA terminology). shadow(d) This selects data that would be shadowed by an antenna of diameter d meters. If d is zero, then the actual diameter of the antennas (if known) is used. If some data is shadowed, it is advisable to use an antenna diameter value greater than the physical antenna size (e.g. 20% larger). auto This selects auto-correlation data only. So the negated form (-auto) selects cross-correlation data only. on This is intended for use with beam switched single dish observations, and selects based on the on variable in the visibility dataset. The on variable indicates whether the telescope is pointing on or off source. bin(n1,n2) This selects visibilities whose pulsar bin number is in the range n1 to n2 inclusive. The current version of SMA software does not support it. or This is discussed below. As noted before, all subcommand names can be abbreviated to the minimum number of characters needed to keep them unambiguous (minimum match). Each subcommand can be prefixed with a plus or minus sign (+ or -). A plus sign means to select the data given by the following subcommand, whereas a minus sign means to discard the data. If neither a plus nor a minus sign is present, a plus sign is assumed. For example select=uvrange(0,10) means to select all visibilities between 0 and 10 kilowavelengths, whereas select=-uvrange(0,10) selects all data except visibilities between 0 and 10 kilowavelengths. Several subcommands can be combined on the same line, separated by commas. When combining several subcommands of different types, the visibility must be selected by all the subcommands to be accepted (a logical AND). When combining several subcommands of the same type, then the visibility is accepted if it is selected after sequentially examining each of the subcommands (a logical OR). For example: select=uvrange(0,10),uvrange(20,30) selects data with a u v radius in the intervals 0 to 10 kilowavelengths as well as 20 to 30 kilowavelengths. The following uses a select then discard approach to selecting the same u v ranges as above: select=uvrange(0,30),-uvrange(10,20) The following selects the same u v ranges, but only for the baseline between antennas 1 and 3.

47 2-30 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS select=uvrange(0,10),uvrange(20,30),antennae(1)(3) The following selects all baselines, with the exception of 1-2, 5-7 and 6-7: select=-antennae(1)(2),-antennae(5,6)(7) Another way of combining subcommands, is with the or subcommand. This allows you to OR together two clauses of selection commands. For example, to select spectral windows 1 and 2 for the times 0:00 to 1:00, and spectral windows 4 and 5 for times 2:00 to 3:00, use: select=time(0:00,1:00),window(1,2),or time(2:00,3:00),window(4,5) By combining the various subcommands and the or subcommand, quite complex selection criteria can be developed. For complex selections, file may be preferred. For example, consider a file, select.data, containing the text: time(1:00,2:00),window(1,2),or time(2:00,3:00),window(1,2),-uvrange(50,100),-antennae(1)(3), or time(3:00,4:00),window(1,2,3,4) Then select=@select.data will use windows 1 and 2 for time 1:00 to 2:00, and windows 1, 2, 3 and 4 for times 3:00 to 4:00. For time 2:00 to 3:00, it uses windows 1 and 2, but omits data for baseline 1-3 which has a u v range of 50 to 100 kilowavelengths. There are a few limitations on the use of amplitude, polarization and window selection. Some of these limitations are mentioned in following sections, whereas others are not. Few of these limitations affect normal practice, and the selection software will inform you if there is a problem. A MIRIAD visibility dataset can contain simultaneous observations at multiple spectral bands, or spectral windows. These are called IF-bands in AIPS. When there are multiple spectral windows, channels are still numbered from 1 to N, where channel 1 is the first channel of the first spectral window, and channel N is the last channel of the last spectral window. That is, channel numbering takes no regard of spectral window boundaries. When there are multiple spectral windows, it is often convenient to be able to select based on the spectral window, rather than a channel range. This is achieved with select=window, which selects data from a set of spectral windows. The general form is select=window(w1,w2,...) where w1, w2, etc, are spectral window numbers. For example, selecting the data from spectral windows 1, 2 and 3 would be achieved with select=window(1,2,3)

48 2.3. UTILITY FUNCTIONS 2-31 Spectral window selection and velocity/felocity linetypes are quite complementary. The different spectral windows might correspond to different spectral lines with different rest frequencies, and so it is desirable to select only a subset of spectral windows (those corresponding to a given spectral line) in forming the output velocity channels. Spectral window selection is not particularly complementary to channel linetype processing both select a range of input channels. When line=channel and select=window are used together, the total apparent number of input channels is equal to the number of channels in the selected spectral windows, and the start channel number of the line parameter is relative to the first selected spectral window. For example, if there are multiple spectral windows, then select=window(2) line=channel,1 would select just the first channel of the second spectral window. Similarly select=window(2) line=channel,10,8,1,2 would select 10 channels, being every second channel starting at the 8th channel of the second spectral window. There are some restrictions on the use of spectral window selection. For channel linetype, the selected spectral windows must be contiguous. For example, you cannot select just windows 1 and 3. This restriction is relaxed by some tasks, most notably uvcat. For channel linetype, window selection cannot be used with or selection. For velocity/felocity linetype, if both window and or selections are used, then there should be a window selection within each or clause. Additionally only one of the or clauses should be satisfiable for a given visibility. For example, consider the following: select=time(0:00,1:00),window(1,2),or time(0:00,2:00),window(3,4) Here data in the times 0:00 to 1:00 could have been selected by either of the two clauses. There is no guarantee as to which of the two window subcommands will be used during this time interval. Having no distinction between single- and multi-source files has the advantage that all visibility tasks can manipulate any visibility dataset. However manipulating a dataset with, say, several sources can require more care. For example, it rarely makes sense to make an image using data from several sources. Generally the user is given reasonable flexibility to perform whatever she/he deems appropriate but this has some disadvantages. For example, consider determining calibration for an observation where the calibrator (a point source) and the program source are within the one dataset. In deriving the antenna gains, one would want to select only the data from the calibrator, whereas when imaging one would select only the data from the program source. The visibility selection mechanism handles this situation. But to forget to select the appropriate data (calibrating with data including the program source, or imaging including the calibrator) would result in a mess. The selection criteria most appropriate for datasets containing multiple sources, frequencies and mosaiced pointing centers are source, frequency, window, ra, dec, dra and ddec. Other examples of using the select keyword are given below:

49 2-32 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS select=window(2) % Select only data from the 2nd IF select=freq(4.74) % Select data where the frequency of the first channel is 4.74 GHz ($\pm 1$\%) select=ra(05:30,06:00) % Select visibilities with RA between 5:50 and 6:00 hours. select=dra(0),ddec(0) % Select data from only the central pointing (delta RA and DEC of 0) of a mosaiced observation. select=source(vela) 5 Take data for source vela only select=window(2),freq(4.74)% Select data from the 2nd IF where its frequency is 4.74 GHz POLARIZATION/STOKES The visibility datasets possibly contain correlations for different polarizations and Stokes parameters. There are two basic ways that polarization characteristics can be measured: firstly, two orthogonal feeds can be present on the antennas, and the four different polarization correlations from a baseline can be taken and recorded. Alternatively only one feed may be present, which can be rotated, or there are not enough correlators to calculate all four polarization correlations simultaneously. The last approach, used by the SMA allows all polarization parameters to be measured in a sequence. In MIRIAD each different polarization/stokes measurement is treated as a separate visibility. So if a baseline measures four simultaneous polarization correlations, then four visibilities will be produced for this baseline, per integration interval. In this case, each visibility will be tagged (with a u v variable pol) to indicate the polarization type. There are two ways the user can determine which polarizations he or she wants to process, either with the select parameter, or with the stokes parameter. The select approach selects visibilities purely on the basis of their polarization or Stokes parameter. It is a normal part of the u v selection process, as described in Section The select mechanism cannot convert from raw polarization parameters to Stokes parameters. It just selects visibilities, in the file, based on their polarization. The general form is: select=polarization(a,b,c,...) where a,b,c etc can be one of the mnemonics For Stokes parameters: i, q, u, v. For circular polarization parameters: rr, ll, rl, lr. For linear polarization parameters: xx, yy, xy and yx. Using the stokes approach is superficially similar, in that it causes only certain polarization/stokes parameters to be processed. However, with the stokes approach, the software can perform conversion between raw polarizations and Stokes parameters if required. It also insists that all the requested polarization/stokes parameters are calculable, at a given time, before it will allow any of them to be processed. For example, if the user requests Stokes I and V, and the dataset contains linear polarization data, then all four of rr, ll, rl and lr polarizations must be present for I and V to be returned in the case of circular polarization. The general form of the stokes approach is:

50 2.4. UTILITY TOOLS 2-33 stokes=a,b,c,... where a,b,c etc, can be one of the mnemonics given above (i, q, u, v, rr, ll etc). The stokes can also take the values ii, qq and uu: Mnemonic ii returns Stokes-I given the assumption that the source is unpolarised. For example MIRIAD would not return any data if you requested stokes=i and the dataset contains only, say, xx (xx is not Stokes-I for a polarized source). How if you requested stokes=ii in this situation, the xx data would be passed through, because you have told MIRIAD that the source is unpolarised. Many tasks use stokes=ii as their default. Mnemonic qq and uu returns Stokes-Q and Stokes-U given the assumption that the parallactic angle is 0. These quantities are rarely of astronomical interest, but they are useful to investigate some instrumental effects. The select mechanism is usually used for time-shared polarization measurements, whereas the stokes mechanism is usually used where there are simultaneous measurements. However this is not a hard and fast rule. The two approaches generally cannot be used at the same time. Check the documentation for each task, especially if both select and stokes parameters are present. 2.4 Utility Tools In this section, we list additional utility tools for inspection and analysis of uv data files INSPECTION OF VARIABLES MIRIAD task, uvcheck, is to check uv-variable values and flag uv-data if variables or data are out of range. Task: uvcheck vis = sgra690_rx2.lsb var = antaz It reports: 0 known problems in this data 0 channels flagged 0 wideband flagged antaz: 8656 values in range: E+21 to E+21 Average= rms= Min= Max= It also can flag the visibility channel data outside a range of the variable specified: Task: uvcheck vis = sgra690_rx2.lsb flagval = flag var = systemp range = 0,8000

51 2-34 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS It reports: 0 known problems in this data 2304 channels flagged 0 wideband flagged systemp: 6489 values in range: to Average= rms= Min= Max= MIRIAD task, uvlist, with options=var,full can list all the variables in a uv file: Task: uvlist vis = sgra690_rx2.lsb options = var,full Header variables at 05FEB16:14:05:13 antaz : antaz : antel : antel : antpos : antpos : antpos : antpos : antpos : calcode :c dec : -5:51: epoch : evector : instrume:sma inttime : ischan : ischan : ischan : jyperk : latitud : 19:49: longitu : 204:31: mount : 0 nants : 8 nchan : 768 npol : 1 nschan : nschan : nschan : nspect : 24 obsdec : -5:53: observer:smauser obsra : 13:10: pol : -5 ra : 13:10: restfreq: restfreq: restfreq: restfreq: restfreq: sdf : E E E E E-03 sdf : E E E E E-03 sdf : E E E E E-03 sdf : E E E E E-03 sdf : E E E E-03 sfreq : sfreq : sfreq :

52 2.4. UTILITY TOOLS 2-35 sfreq : sfreq : source :callisto sourid : 1 systemp : systemp : tcorr : 0 telescop:sma veldop : veltype :VELO-LSR version :test vsource : BASELINE FITTING bee is a Miriad task to fit polynomials and antenna positions to the antenna gains. The gains can be plotted versus time, HA, DEC, elevation, etc, and identified by source. Polynomial or cursor drawn fits to the amplitude and phase can be written as antenna gains and used to calibrate the data. Here is a procedure to do the baseline fitting: 1. Selfcal for phase: Task: selfcal vis = xmiriad050218_230_rx0.lsb select = source(ti*,3c111*,ga*,ce*,17*) interval =.5 refant = 6 offset = line = chan,1,1, Start bee: Task: bee vis = miriad050128_230_rx0.lsb device = /xs Here are the interactive instruction for using the fitting, plotting and editting functions that are supported in bee: Number of points= 130 Enter antenna to display : 1 Number of points= 130 AVAILABLE COMMANDS ARE: 1P Fit axis offset 2P 2-parameter fit [sets B3 = -cotan(lat)*b1] 3P 3-parameter fit to phase differences 4P 4-parameter fit; no 2PI discontinuities! 5P 5-parameter fit includes axis offset AM Plot amplitudes versus time ED Edit to new baseline etc. EX Exit BEE and write antpos file FI Seach for best fit baseline GR Plot vs. HA, sin(dec), elevation, or focus IN Get the Gains for another antenna LI List antenna positions and current fit.

53 2-36 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS PH Plot phase versus time TE Edit data to test fitting programs TF Tpower and Tair fit to phase TH Plot airtemp versus time TP Plot tpower versus phase and fit slope. WA Write ANTPOS file. WG Write gains into the output gains file. 3. Plotting Phase: One can use one of the commands, such as PH for plotting phase vs. time. Here is an exmaple: command=ph Plot phase versus time and fit. SELECT SOURCES Type source numbers separated by commas <cr> means finished; <a> means all sources 1 titan 2 3c111 3 ganymede 4 ceres SOURCES: a in plot window type CURSOR OPTIONS: A restore deleted points B set bottom to cursor position C Connect cursor positions D move phase point down by 2pi E end manipulations on plot F show instrumental phase fit G fit gains to tpower plot H Hanning smooth fitted curve I Identify the sources J Join the plotted points K superpose temperature plot L set left to cursor position M plot more sources N set new limits and replot O superpose tpower plot P replot on printer Q position fit R set right to cursor position S replot T set top to cursor position U move phase point up by 2pi V delete point(s) W write the source labels at top of plot X average and rms of plotted points 0-8 order of polynomial fit + unwrap phase 4. Baseline Fitting: bee also supports least square fitting to the phase difference for new antenna positions: command=in % choose new antenna

54 2.4. UTILITY TOOLS 2-37 Figure 2.11: The phase of all the calibrators (for antenna 1) as function of time is plotted using MIRIAD bee.

55 2-38 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Enter antenna to display :1 % choose anetnna 1 Number of points= 845 command=4p % 4parameter fits 4-PARAMETER FIT Fit antenna positions and constant phase offset SELECT SOURCES Type source numbers separated by commas <cr> means finished; <a> means all sources 1 titan 2 3c111 3 ganymede 4 ceres SOURCES: a % average all source >>>> report of the results from the baseline fit to antenna 1 <<<< 4 parameter fit, sigma: initial baseline new baseline delta baseline uncertainties covariance matrix command=fi Step size for bx/by/bz ( ) : # of steps in 1-D ( 40.) :15 Step size for antmiss ( ) : # of steps (1=nofit) ( 1.) :1 BEE FIT DATA--starting with: bx = by = bz = antmiss = chisq = 3.09 bx = by = bz = antmiss = chisq = 1.63 bx = by = bz = antmiss = chisq = 1.63 bx = by = bz = antmiss = chisq = 1.63 bx = by = bz = antmiss = chisq = 1.62 bx = by = bz = antmiss = chisq = 1.62 bx = by = bz = antmiss = chisq = 1.53 bx = by = bz = antmiss = chisq = 1.45 bx = by = bz = antmiss = chisq = 1.44 bx = by = bz = antmiss = chisq = 1.44 bx = by = bz = antmiss = chisq = 1.39 bx = by = bz = antmiss = chisq = 1.32 bx = by = bz = antmiss = chisq = 1.32 bx = by = bz = antmiss = chisq = 1.19 bx = by = bz = antmiss = chisq = 1.10 bx = by = bz = antmiss = chisq = 1.10 bx = by = bz = antmiss = chisq = 1.02 bx = by = bz = antmiss = chisq = 1.01 bx = by = bz = antmiss = chisq = 1.01 bx = by = bz = antmiss = chisq = 1.01 bx = by = bz = antmiss = chisq = 1.01 bx = by = bz = antmiss = chisq = 1.01 command=in Enter antenna to display :2 % for antenna 2

56 2.4. UTILITY TOOLS 2-39 Number of points= 846 command=4p 4-PARAMETER FIT Fit antenna positions and constant phase offset SELECT SOURCES Type source numbers separated by commas <cr> means finished; <a> means all sources 1 titan 2 3c111 3 ganymede 4 ceres SOURCES: a 4 parameter fit, sigma: initial baseline new baseline delta baseline uncertainties covariance matrix command=fi Step size for bx/by/bz ( ) : # of steps in 1-D ( 40.) :15 Step size for antmiss ( ) : # of steps (1=nofit) ( 1.) :1 BEE FIT DATA--starting with: bx = by = bz = antmiss = chisq = 0.10 bx = by = bz = antmiss = chisq = 0.10 bx = by = bz = antmiss = chisq = 0.10 bx = by = bz = antmiss = chisq = 0.10 One can go through the least square fitting for all the antennas to obtain corrections for positions. The gains derived from amplitude and phase fits are write to the gains file and the new antenna position can be written out to a text file. Here is an exmaple: command=ex The gains derived from amplitude and phase fits using the ED command are written at this time. The gains are written into the output gains file. The default output is the input uvdata file. OK to write the gains?y Number of solution intervals: Write New Antenna Positions: Enter filename for antenna positions: antpos.new #Fitted antenna positions - Mon Mar 21 16:25:

57 2-40 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS LINEAR REGRESSION BETWEEN 230/690 GHz PHASES varfit is to analyse the correlations of antenna gains and uv-variables. varfit fits the following: yaxis = slope * xaxis + intercept D(yaxis) = slope * D(xaxis) + intercept where D() is the difference between succesive samples, xaxis, yaxis are the uv variables from single uvfiles. In addition, varfit provides an option of regression for the phases derived from the gains of two uvfiles. If either axis is an antenna gain, then selfcal must be run first and the axes are sampled at the interval used in selfcal. For linear regression between the two phase solutions, the setup in the selfcal for the two simultaneously sampled data files needs identical. The rms and correlation coefficient are listed for each antenna. varfit is a useful tool for analysis of the data obtained from observations at 230/690 GHz, simultaneously. An exmaple is shown to do linear regression between the phases derived from 230 and 690 GHz data sets based on the observations of Ceres on Feb 18, 2005 during the 690 GHz campaign. Ceres, 0.5 arcsec in size, was observed for 3.5 hrs at both lsb and usb. Here is the procedure to extract data from the raw SMA archival data: 1. Loading data: \rm -r miriad050218_230_rx0.lsb smalod in=/home/miriad/smadata/050218_04:23:14/ out=miriad050218_230 rxif=0 \ sideband=0 options=dospc rsnchan=, \ nscans=0, \rm -r miriad050218_690_rx2.lsb smalod in=/home/miriad/smadata/050218_04:23:14/ out=miriad050218_690 rxif=2 \ sideband=0 options=dospc rsnchan=, \ nscans=0, 2. Qvack the first integrations after switching source: qvack vis=miriad050218_230_rx0.lsb interval=0.5 mode=source qvack vis=miriad050218_690_rx2.lsb interval=0.5 mode=source 3. Uvsplit ceres data from the multi-source files: uvsplit vis=miriad050218_230_rx0.lsb select= source(ceres) options=nowindow uvsplit vis=miriad050218_690_rx2.lsb select= source(ceres) options=nowindow rename the file names: mv ceres ceres.221lsb mv ceres ceres.682lsb 4. Uvplt the two data sets together:

58 2.4. UTILITY TOOLS 2-41 uvplt vis=ceres.221lsb,ceres.682lsb axis=time,pha options=nocal,unwrap \ device=/xs nxy=3,5 5. Flag problematic data due to instruments: There is a large phase drift on the baselines related to antenna 4 in the last 2.5 hrs of the observation at 221 GHz, which is apparently due to the instrument and mars the linear correlation between 221 GHz and 682 phases. To make the data sampling in the two files (221/682) identical, we flag all the antennas in the last 2.5 hrs for both files. uvflag vis=ceres.221lsb select= time(17:27,23:00) flagval=flag uvflag vis=ceres.682lsb select= time(17:27,23:00) flagval=flag 6. Selfcal to derive antenna-based phase: selfcal vis=ceres.221lsb interval=3 refant=6 line=chan,1,1,3072 selfcal vis=ceres.682lsb interval=3 refant=6 line=chan,1,1, Varfit for linear regression of the 221/682 phases: Input: Task: varfit vis = ceres.221lsb,ceres.682lsb xaxis = time yaxis = phase log = device = /xs nxy = 2,3 xrange = yrange = refant = 6 refant2 = options = phareg Report on screen: Number of points= 24 Telescope: SMA Longitude: Latitude: yaxis = phase2 derived from gains of FILE1: ceres.682lsb xaxis = phase1 derived from gains of FILE2: ceres.221lsb yaxis = slope * xaxis + intercept ant yaxis_ave yaxis_rms slope intercept rms-fit correlation D(yaxis) = slope * D(xaxis) + intercept

59 2-42 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.12: Phase1 at 221 GHz (black dots) and phase2 at 682 GHz (red dots) versus time are plotted for each of the baselines.

60 2.4. UTILITY TOOLS 2-43 ant yaxis_ave yaxis_rms slope intercept rms-fit correlation Plots the linear regression of 690/230 phases: Plots the variation of the residual phase: PHASE RESIDUAL VS. AMBIENT TEMPERATURE One might want to further investigate the origin of the variations in the residual phase. One possiblity is that changes of ambient temperature might affect on the residual phase. The ambient temperature data along with other engineer data has been stored in SMA sybase. Here we provide a C-shell script to extract ambient temperature data from the SMA Sybase TambSybase.csh. The script (TambSybase.csh) can be also obtained from $MIR/examples. On the linux computer (d2o.sma.hawaii.edu) at the SMA site in Hawaii, one can execute this script. There are different type of tempratures that were recorded and stored as a RM variable in SMA Sybase. For example, RM WEATHER TEMP F is the weather temperature outside of the antenna cabins, and RM AMBIENTLOAD TEMPERATURE F is the best measure of the temperature inside of each antenna cabin. Here is the usage for extracting the weather temperature from SMA Sybase: d2o% TambSybase.csh RM\_WEATHER\_TEMP\_F The arguments are year, month, day and name of a RM variable. An ASCII data file (tambient.dat) is created. Then copy the data file to one s miriad working area: scp tambient.dat miriad@tornado.cfa.harvard.edu:/data/miriadwork/ Input: Task: varfit vis = ceres.221lsb,ceres.682lsb xaxis = time yaxis = phase device = /xs nxy = 2,3 refant = 6 options = phareg,tambient The options=phareg,tambient in MIRIAD varfit allows users to do the linear correlation between the residual phase and the ambient temperature and plot the ambient temperature as a function of UT time. Here is a report on the screen: Ph_rsd(yaxis) = slope * T_amb(xaxis) + intercept ant yaxis_ave yaxis_rms slope intercept rms-fit correlation

61 2-44 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.13: The linear regression between phase1 (221 GHz) and phase2 (682 GHz) derived from MIRIAD varfit for each antenna. Antenna 6 is the reference antenna.

62 2.4. UTILITY TOOLS 2-45 Figure 2.14: The residual phase (phase slope*phase intercept) as function of time calculated from varfit for each antenna. Antenna 6 is the reference antenna.

63 2-46 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS The correlation coefficients are small for all the antennas, suggesting that the residual phase does not relate to the variation of the weather temperature (Fig2.15): GLUE SMA MULTI-CHUNK FILES smachunkglue is to glue together individual multi-chunk files into one mega-multi-chunk file. This progarm is useful for IDL users who calibrate SMA data by breaking an SMA 2GHz spectrum in several pieces and would like to glue the individuals together back to a single file for further imaging and analysis in Miriad. Here is the usage for gluing two files (ngc3079-lmay9 1, ngc3079-lmay9 2): Task: smachunkglue vis = ngc3079-lmay9 nfiles = 2 out = ngc3079-lmay9glued TOOL FOR OBSERVING SCHEDULE schedobs is newly implemented in MIRIAD for users to graphically schedule observations of sources. For given observing date, beginning of observing time and observing interval, schedobs simulates the tracks of the sources in terms of their elevations as functions of time. Here is the usage: Task: schedobs axis = time,el %time vs. elevation obsdate = 2006mar15 %observing date UTstart = 10,30,0 %beginning time in UT obshours = 15 %observing duration in hours sfile = vlbisource.txt %text file of the source list device = /xs %graphical device nxy = 1,1 %1 plot per page options = hst %convert to HST mirhome = $MIR %the path of Miriad home Here is a template of the input text file of the source list: 3c279 12:56: :47:21.52 callisto sgrastar 17:45: :00: :24: :14: c :53: :08:53.56 The source name must be given in 8 characters or less as required by the MIRIAD convention SORTING UVDATA INTO A TIME ORDER In the case of merging several uvdata sets (output from IDL-MIR) using either uvcat or uvaver, the resulted uvdata might be not in the ascending time order. The calibration tasks in MIRIAD might reject processing such an irregular dataset since in the MIRIAD convention the uvdata must be sorted into ascending order in time. MIRIAD task uvsort copies a uv dataset, sorting uvdata in a random time order into ascending order in time. Here is the usage:

64 2.4. UTILITY TOOLS 2-47 Figure 2.15: The weather temperature as function of time plotted with MIRIAD varfit for each antenna. Antenna 6 is the reference antenna. The weather temperature data were extracted from SMA Sybase corresponding to the observations on

65 2-48 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS Figure 2.16: The tracks of the sources scheduled to be observed on 2006Mar15 is simulated with schedobs in MIRIAD Figure 2.16: The tracks of the sources scheduled to be observed on 2006Mar15 is simulated with schedobs in MIRIAD uvsort vis=/data/smadata/dmarrone/bp.6.usb \ out=bp.6.usb.tsrt One can check up the result with uvindex. Summary listing for data-set /data/smadata/dmarrone/bp.6.usb Time Source Antennas Spectral Wideband Freq Record Name Channels Channels Config No. 05JUL30:11:21:34.5 URANUS JUL30:13:21:51.5 URANUS JUL30:11:14:23.1 NEPTUNE JUL30:13:19:37.4 NEPTUNE JUL30:08:18:14.2 3C JUL30:16:12:22.7 Total number of records Total observing time is 2.87 hours The input data-set contains the following frequency configurations: Frequency Configuration 1 Spectral Channels Freq(chan=1) Increment GHz GHz

66 2.4. UTILITY TOOLS GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz The input data-set contains the following polarizations: There were 4038 records of polarization LR There were 4048 records of polarization RL There were 6640 records of polarization LL There were 6305 records of polarization RR The input data-set contains the following pointings: Source RA DEC dra(arcsec) ddec(arcsec) 3C :53: :08: NEPTUNE 21:15: :02: URANUS 22:47: :33: Summary listing for data-set BP.6.usb.tsrt Time Source Antennas Spectral Wideband Freq Record Name Channels Channels Config No. 05JUL30:08:18:14.2 3C JUL30:11:14:23.1 NEPTUNE JUL30:11:21:34.5 URANUS JUL30:11:26:42.1 3C JUL30:11:35:28.4 NEPTUNE JUL30:11:37:44.8 URANUS JUL30:11:40:04.5 3C JUL30:11:48:51.3 NEPTUNE

67 2-50 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS 05JUL30:11:51:08.0 URANUS JUL30:11:53:27.4 3C JUL30:12:09:42.7 NEPTUNE JUL30:12:11:59.1 URANUS JUL30:12:14:35.0 3C JUL30:13:19:37.4 NEPTUNE JUL30:13:21:51.5 URANUS JUL30:13:24:09.5 3C JUL30:13:32:49.6 NEPTUNE JUL30:13:40:47.3 URANUS JUL30:13:43:04.2 3C JUL30:13:56:14.0 NEPTUNE JUL30:13:58:54.5 3C JUL30:14:09:25.8 URANUS JUL30:14:23:22.0 3C JUL30:14:27:59.1 NEPTUNE JUL30:14:30:36.3 URANUS JUL30:14:32:52.7 3C JUL30:14:45:57.0 URANUS JUL30:14:48:12.1 3C JUL30:15:01:16.0 URANUS JUL30:15:03:31.1 3C JUL30:15:16:31.4 URANUS JUL30:15:18:44.3 3C JUL30:15:31:42.5 URANUS JUL30:15:33:55.5 3C JUL30:15:46:52.8 URANUS JUL30:15:49:06.5 3C JUL30:16:02:03.5 URANUS JUL30:16:04:17.6 3C JUL30:16:12:22.7 Total number of records Total observing time is 2.87 hours The input data-set contains the following frequency configurations: Frequency Configuration 1 Spectral Channels Freq(chan=1) Increment GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz

68 2.4. UTILITY TOOLS GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz GHz The input data-set contains the following polarizations: There were 4038 records of polarization LR There were 4048 records of polarization RL There were 6640 records of polarization LL There were 6305 records of polarization RR The input data-set contains the following pointings: Source RA DEC dra(arcsec) ddec(arcsec) 3C :53: :08: NEPTUNE 21:15: :02: URANUS 22:47: :33: Except for the first section, other sections between the orignal dataset and the sorted dataset should be the same.

69 2-52 CHAPTER 2. UV DATA LOADING, DISPLAY, FLAGGING AND CORRECTIONS

70 Chapter 3 Strategies in Calibrations 3.1 Introduction In MIRIAD a model of gains to the visibility data can be described as g(t)g(ν)exp [i2πτ(t)(ν ν 0 )] where g(t) is a time variable complex number (or called antenna gain), g(ν) is the bandpass function, ν is frequency and τ is a time-varying delay term. There are different approaches in solving for gains depending on the requirements by individual programs. In this Chapter, a general approach in the gain calibrations is given. In addition, strategies in calibrations for different science projects are being considered and are (will be) supported by the SMA software group. 3.2 SMA Spectral Data and Calibration Strategy The SMA correlator can flexibly change its configuration and produce spectral data with a combination of multiple spectral resolutions and multiple bandwidths in the multiple spectral chunks (windows). It is often that the SMA data is taken from a single polarized antenna feed. The SMA correlator samples the data from individual antennas and correlates the data from each of an antenna pairs (also called baseline). A large data stream with multiple spectral windows (or chunks) in each of the upper and lower side bands are produced for a short integration time. For non-polarization observations, the solutions of the antenna gains (complex or in both amplitude and phase), bandpass (normalized bandpass shape in both amplitude and phase) and the delay term can be solved directly from the observations of calibrators. A good calibrator must be strong with a simple emission structure. A strong point source is an ideal calibrator BANDPASS AND GAIN SOLVERS The MIRIAD task mfcal determines antenna gains, band passes, and delay parameters from a calibrator for a multiple-frequency observations. A multiple-source observation with dual receivers is supported. mfcal is an excellent program for calibration of antenna gains and bandpass if one can find compact (pointlike) and strong calibrators. At submillimeter wavelengths, quasars become weak and phase instability appears to prevent from gaining S/N by simply integrating bandpass data. Furthermore, the location of the nulls in the visibilities of a resolved planet disk (in particular, those large planets) is a function of frequency 3-1

71 3-2 CHAPTER 3. STRATEGIES IN CALIBRATIONS in addition to the phase flip across a null. The complex in the visibility of planets might have lead to artifacts in the bandpass solutions calculated from software program which only good for the sources with simple visibility structure, such as a point source. Software could be imperfect in handling complicated visibility data. To cope with the specific problems in calibrations at submillimeter wavelengths, on the basis of mfcal, a MIRIAD task, smamfcal, has been developed for enhancing the reduction of SMA data. A few pre-processing algorithms such as moving smooth, vector normalization, and amplitude-weighting prior to solving for bandpass and/or antenna gains have been implemented in smamfcal. In addition, due to some instrumental effects, slow bandpass variations in time on some SMA antennas were seen. Such a variation could cause ripple across the spectral band and damage bandpass quality. The time variation effect in bandpass need to be removed, in particular, for weak line and strong continuum emission cases which require a high spectral dynamic range. A special tool smatbpass has been supported in MIRIAD to correct for bandpass ripples. The SMA correlator consists of a large number of chips which compute the chunk of spectral visibility for baselines (antenna pairs). Occationally, failures of computer chips or errors in handling the data were showed. Those errors due to the correlator are usually baseline based. The baseline-based errors can be corrected using blcal in MIRIAD We demonstrate calibration procedures for both the normal and special cases in the following subsections: 3.3 Normal Calibration Procedure BANDPASS Normally, an observation track includes a bandpass scan on a small planet or a strong quasar. As long as the visibility data on each channel gives adequate S/N, one can solve for antenna-based bandpass solutions using smamfcal. Here is an exmaple for solving for bandpass based on observations of the Jupiter s moon, Callisto. Task: smamfcal vis = gc_rx1.lsb.tsys % input data with Tsys correction and bad data that are flagged. select = source(call*) % here select a calibrator with wildcard call* = callisto refant = 3 interval = 100 weight = 2 % time interval for bandpass solutions % the channel visibility is normalized % by the average of channels specified % by the inner 75% of each "spectral chunk"; The above setup of smamfcal for bandpass is given using the calibrator Callisto. The solutions can be checked with smagpplt: smagpplt% inp Task: smagpplt vis = gc_rx1.lsb.tsys % input file with bandpass table device = /xs % x-window device log = yaxis = amp,phase % displays amplitude and phase

72 3.3. NORMAL CALIBRATION PROCEDURE 3-3 options = bandpass,opolyfit % bandpass; bandpass solutions will be replaced by the polynomial fit polyfit = 5 % the 5th order polynomial is used in the least square fit. nxy = 2,4 % page setup: 2 in row and 4 in column. The solutions for amplitude and phase are shown in (Figs. 3.1 and 3.2). The ripples in amplitude shown in the antenna based solutions across the 2 GHz band needs to be corrected. The data of antenna 2 is relatively noisy and poor. Antenna 2 has been removed before solving for bandpass. Antenna 4 shows a large delay across all the chunks. Antenna 6 shows a big phase jump between the first 3 blocks (12 chunks) and the next 3 blocks. The phase solutions of the antennas are with respect to the reference antenna 3. The errors must be removed. This is particularly important for continuum data which is an average of the spectral data. Also the bandpass solution appears to be little noisy and one can fit a polynomial to the solutions. The solid curves are the fitted version of the bandpass solutions. If the polynomial fit is satified, one can run smagpplt by re-setting the options to options = bandpass,opolyfit with other parameters remaining the same. Then the bandpass solutions are replaced by the polynomial fit. Note that the options of opolyfit will replace the original solution. We encourage users to try these parameters. If mistakes occur, one can always re-run smamfcal to calculate the solutions again ANTENNA GAINS We now process to calculations of the antenna gains using a phase calibrator. In this testing program, NRAO 530 (a quasar) was used in the observation. This source is a point source with flux density about Jy at 335 GHz. We use smamfcal to solve for the antenna gains. A new keyword weight=1 is set, which allows to weight the visibility by 1/σ 2, where σ is the r.m.s. noise of the visibility data. The solution interval of 1 minutes (which is smaller than the scan time of 3 min). The solutions of the antenna gains can be smoothed or polynomial interpolated afterwards. Here is a setup: smamfcal %inp Task: smamfcal vis = gc_rx1.lsb.tsys select = source(nra*) % select nrao530 flux = 2 % flux density assumed refant = 3 % reference antenna interval = 1 % solution interval weight = 1 % weight by 1/sigma**2 options = nopassol % not solve for bandpass The solutions of antenna gains can be inspected with smagpplt. A setup for smagpplt is shown below. There may be a few bad points in the gain solutions. One may need to go back to smablflag to do further data flagging. Or, one can use gplist and gpedit to reset bad point to (amp,phase)=(1,0). Alternatively, smagpplt provides a nice feature for smoothing or polynomial fitting to the antenna gains. An intermediate order of polynomial function may be useful to fit the gain solutions in order to reduce the noise in the gains. Figs. 3.3 and 3.4 show the plots of antenna-based gain solutions (red points) in real and imaginary fit with the 5th order polynomial curves (blue). One can replace the original gain solutions with the polynomial fit by adding opolyfit to options or options = gains,opolyfit. Note that in this observation, antenna 2 showed odd values of the gain solutions in comparison with other antennas. The problematic antenna may contaminate the data. The antenna 2 has been flagged out. One may get rid of the data related to problematic antennas prior to further data processing. In most cases, the data from low S/N antennas will mar the quality of images. If the baseline coverage is not a concern, it would be wise to remove the bad quality data immediately.

73 3-4 CHAPTER 3. STRATEGIES IN CALIBRATIONS Figure 3.1: solutions. Bandpass solutions in amplitude. The solid black curves are the 5th polynomial fit to the

74 3.3. NORMAL CALIBRATION PROCEDURE 3-5 Figure 3.2: Bandpass solutions in phase. The solid black curves are the 5th polynomial fit to the solutions.

75 3-6 CHAPTER 3. STRATEGIES IN CALIBRATIONS smagpplt% inp Task: smagpplt vis = gc_rx1.lsb.tsys device = /xs yaxis = real,imag options = gains,opolyfit polyfit = 5 nxy = 2, APPLY BANDPASS AND ANTENNA GAINS So far, we have not corrected the visibility data. The flags, bandpass and antenna gain solutions which are stored in the numerous tables (flags, bandpass, and gains) under the uv file directory. If one is happy with these solutions, they can be applied to the visibility data using uvaver by the setup below: uvaver% inp Task: uvaver vis = gc_rx1.lsb.tsys select = source(sgra*, sgrb*, nrao530) out = gc_rx1.lsb.tsys.cal The data gc rx1.lsb.tsys.cal is the calibrated data. With uvaver, one can average the data from each chunk to produce a ch0 data set. The corresponding setup is given below: uvaver% inp Task: uvaver vis = gc_rx1.lsb.tsys.cal line = channel, 24,2,13,16 out = gc_rx1.lsb.tsys.cal.ch0 To make a continuum data, the setup below is suggested: uvaver% inp Task: uvaver vis = gc_rx1.lsb.tsys.cal.ch0 line = channel, 1,1,24 out = gc_rx1.lsb.tsys.cal.cont If the source has sufficient S/N for selfcal, it would be better to stay with the chunk-based ch-0 data set. Then, the option of MFS in the self-calibration and imaging process will minimize the residual errors before averaging. One can break the multi-source file into single source files using uvsplit uvsplit% inp Task: uvsplit vis = gc_rx1.lsb.tsys.cal options = nowindow

76 3.3. NORMAL CALIBRATION PROCEDURE 3-7 Figure 3.3: Antenna gain solutions in real (red dots) fitted with the 5th order polynomial (blue curves).

77 3-8 CHAPTER 3. STRATEGIES IN CALIBRATIONS Figure 3.4: curves). Antenna gain solutions in imaginary (red dots) fitted with the 5th order polynomial (blue

78 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-9 As we have shown, the unwrapped phase across the broad SMA band (2 GHz) shows a large drift for some antennas (particularly for the antennas with long baselines). Even for continuum imaging, the bandpass calibration will be crucial for removing the phase ripples to increase the dynamic range. The calibration feature for multiple frequency windows (or chunks) in MIRIAD tasks mfcal or smamfcal is highlighted for synthesizing a continuum image from a large number of spectral windows s data. 3.4 Special Calibration Tools and Procedures The normal calibration procedures are discussed above. Sophiscated users might use the advanced tools in MIRIAD to correct for some unusual errors under certain circumstance BANDPASS RIPPLES After bandpass calibration following the normal procedure discussed above, one might need to pay attention to the stability of the bandpass in time if significant ripples remained across the 2 GHz band. Identify Bandpass Ripples If several bandpass scans were included during an observing track, one could solve for the antenna-based bandpass at each scan time and then compare the solutions to check up if the bandpass on which antennas varies with time. Here is an exmaple of the observation carried out on August 1, Three data scans from the time intervals of 04:50 to 05:10 on Callisto, 08:00 to 10:15 on 3c454.3 and 10:15 to 15:00 3c454.3 can be separated using either uvsplit or uvaver: uvaver vis=sgrb228a_rx0.lsb.tsys select= source(callisto) \ out=05aug1bp_1 options=nocal,nopass uvaver vis=sgrb228a_rx0.lsb.tsys select= source(3c454.3), time(08:00,10:15) \ out=05aug1bp_2 options=nocal,nopass uvaver vis=sgrb228a_rx0.lsb.tsys select= source(3c454.3), time(10:15,15:00) \ out=05aug1bp_3 options=nocal,nopass Then solve for antenna-based bandpass for each of the time intervals using smamfcal: smamfcal vis=05aug1bp_1 edge=5,5 \ weight=2 refant=3 interval=600 smamfcal vis=05aug1bp_2 edge=5,5 \ weight=2 refant=3 interval=600 smamfcal vis=05aug1bp_3 edge=5,5 \ weight=2 refant=3 interval=600 Now, one could check up whether the bandpass varies by calculating the ratios of the bandpass solutions derived from the time intervals at different time with smagpplt: smagpplt vis=05aug1bp_3,05aug1bp_2 device=/xs \ yaxis=amp \ options=bandpass,wrap,ratio polyfit=5 \ nxy=2,4 yrange=0.7,1.3 sleep 1

79 3-10 CHAPTER 3. STRATEGIES IN CALIBRATIONS smagpplt vis=05aug1bp_3,05aug1bp_2 device==/xs \ yaxis=pha \ options=bandpass,wrap,ratio polyfit=5 \ nxy=2,4 yrange=-20,20 The ripple on antenna 1 shows large variation across the 2 GHz bands in both amplitude ( 20%, see Fig. 3.5) and phase (15 degree, see Fig. 3.6). Correct for Bandpass Ripples The MIRIAD task smatbpass removes time-dependent bandpass ripples. smatbpass reads multiple files (bpfile) containing antenna-based bandpass solved using either smamfcal or mfcal in time intervals at different time. The bandpasses are interpolated or extrapolated to a time with either linear fit from the two nearby solution points or orthogonal polynomial fit from all the solution points in the input bpfiles. Here is a usage of smatbpass: Task: smatbpass bpfile = 05aug1bp nfiles = 3 % root name of input files each of which contains bandpass solutions in the bandpass table at the corresponding time in the gain table. % The number of bandpass files (bpfile) to read. vis = sgrb228a_rx0.lsb.tsys % The name of the input uv data sets. bptime = 30 % The time interval, in minutes, to interpolate/extrapolate the bandpass solutions. options = nocal out % extra options = sgrb228a_rx0.lsb.tsys.bp % The name of the output uv data set which has been applied for bandpass correction. The visibility data in output file sgrb228a rx0.lsb.tsys.bp has been corrected for the bandpass ripples BANDPASS CORRECTIONS ACROSS SIDEBANDS In some cases, one of the sidebands is affected by the pressure-broadened lines in the Earth atmosphere (e.g. the water and ozone lines). Users might want to try corrections for the instrumental bandpass on the data using the bandpass solutions derived from the image sideband. MIRIAD smagpplt has provided an option xpass to transfer the bandpass solutions from the image sideband. Assuming vis rx1.lsb.tsys and vis rx1.usb.tsys, the multiple source files of the lower-side-band and upper-side-band with Tsys corrections, here is a recipe for users to try the bandpass corrections across sidebands in MIRIAD 1. Create small vis-files with the bandpass calibrator uvaver vis=vis_rx1.lsb.tsys select= source(3c279) out=340bp_lsb \ options=nopass,nocal uvaver vis=vis_rx1.usb.tsys select= source(3c279) out=340bp_usb \

80 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-11 Figure 3.5: The amplitude ratio of bandpass solutions derived from two different intervals of time(08:00,10:15) and time(10:15,15:00). The ripple across the 2 GHz band on antenna 1 (left-top panel) shows large variaition in amplitude ((max-min)/mean 20%).

81 3-12 CHAPTER 3. STRATEGIES IN CALIBRATIONS Figure 3.6: The phase ratio of bandpass solutions derived from two different interval of time(08:00,10:15) and time(10:15,15:00). The ripple across the 2 GHz band on antenna 1 (left-top panel) shows large variaition in phase ((max-min)/mean 15 degree).

82 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-13 options=nopass,nocal 2. Derive bandpass solutions smamfcal vis=340bp_lsb \ refant=5 interval=1000 edge=20,20 \ select= pol(rr,ll),window(12,13,14,15) \ weight=2 options=msmooth,averrll smooth=5 smamfcal vis=340bp_usb \ refant=5 interval=1000 edge=20,20 \ select= pol(rr,ll),window(12,13,14,15) \ weight=2 options=msmooth,averrll smooth=5 3. Plot bandpass solutions smagpplt vis=340bp_lsb device=1/xs yaxis=a,p \ options=bandpass \ polyfit=4 nxy=2,3 $< smagpplt vis=340bp_usb device=1/xs yaxis=a,p \ options=bandpass \ polyfit=4 nxy=2,3 4. Transfer the bandpass solutions across sidebands (lsb to usb) smagpplt vis=340bp_lsb,340bp_usb \ options=xpass 5. Copy over the bandpass to the target file gpcopy vis=340bp_usb out=vis_rx1.usb.tsys 6. Apply the bandpass uvaver vis=vis_rx1.usb.tsys out=vis_rx1.usb.tsys.bp BASELINE-BASED BANDPASS Identify Baseline-based Bandpass Errors One might need to inspect the spectra on each baseline in the uv data set (sgrb228a rx0.lsb.tsys.bp) after the antenna-based ripples are corrected. We pick up the strong point source 3c454.3 to identify any baselinebased errors in the bandpass: smauvspec vis=sgrb228a_rx0.lsb.tsys.bp \ select= source(3c454.3) interval=1000 \ axis=freq,both device=/xs nxy=1,1 yrange=0,40 Figs. 3.7 and 3.8 display some of the baselines which clearly show large chunk-based offsets in amplitude on some baselines (e.g. baselines 1-4 for chunk 23 and 3-4 for chunk 15.) or significant phase drift along the channels on some other baselines (e.g. baselines 2-7 and 4-7). The baseline-based errors must be removed prior to the further data reduction.

83 3-14 CHAPTER 3. STRATEGIES IN CALIBRATIONS Figure 3.7: Examples of spectra after correction for bandpass ripples show the residual baseline-based errors, such as the large chunk offsets in amplitude on chunk 23 of baseline 1-4 and chunk 15 of baseline 3-4.

84 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-15 Figure 3.8: Examples of spectra after correction for bandpass ripples show the residual baseline-based errors, such as the large phase drift along channel on baselines 4-7 and 2-7.

85 3-16 CHAPTER 3. STRATEGIES IN CALIBRATIONS Correct for Baseline-based Bandpass Errors In order to correct for the baseline-based bandpass errors, the visibility data of the strong QSO 3c454.3 needs to be extracted from the uv data applied for antenna-based bandpass correction ( e.g. sgrb228a rx0.lsb.tsys.bp) using uvaver: uvaver vis=sgrb228a_rx0.lsb.tsys.bp select= source(3c454.3) \ out=05aug1bl options=nocal,nopass Then, using blcal, the baseline-based correction can be drived from 05aug1bl and be applied to sgrb228a rx0.lsb.tsys.bp: blcal vis=05aug1bl,sgrb228a_rx0.lsb.tsys.bp \ interval=5000 out=sgrb228a_rx0.lsb.tsys.bp.bl The output file sgrb228a rx0.lsb.tsys.bp.bl is the final uv data set which has been applied for the bandpass corrections. Note that blcal assumes a point-source model in the process of corrections for the baselinebased bandpass, ı.e. any deviation from point source model of the source in the first input data (05aug1bl) might introduce errors. In addition, the point source used must be strong enough for calculating bandpass corrections with adequate S/N on each of the baselines. After bandpass calibration, one might check up the spectra for one of the target sources using smauvspec: smauvspec vis=sgrb228a_rx0.lsb.tsys.bp.bl \ select= source(sgrb2_m),ant(3)(4) \ axis=freq,both device=/xs nxy=1,1 Fig. 3.9 shows the spectrum of baseline 3-4 for SgrB2 M. Note that the data in the uv data set sgrb228a rx0.lsb.tsys.bp.bl output from blcal has been applied for all the bandpass corrections and there should no bandpass table attached to this data file. From this point, one can go back to smamfcal for antenna gain corrections. Handling for Baseline-based Bandpass Errors with Weaker/Resolved Calibrators The lack of strong point source for bandpass calibration is the well known problem. Often we run out of the signal budget in baseline-based bandpass calibration. The problem gets worse at the short sub-millimeter wavelengths. Although the planets and the Moon have strong dust emission, the structure of the disk might ruin the closure relation. MIRIAD smablsrc can turn a planet or the rim of the Moon to a pseudo-point source for the purpose of baseline-based bandpass calibration. Here is a usage: smablsrc vis=uvdata \ select= source(mars) \ out=marssblsrc The output file marssblsrc contains the uv data of the pseudo-point source, which can be used for baselinebased bandpass calibration using blcal. In addition, the phase fluctuation are naturally canceled out in the normalization process. One can process the data of weaker point sources using smablsrc prior to integrating the visibility data together to improve S/N for the purpose of baseline-based bandpass calibration: smablsrc vis=uvdata \ select= source(3c111, ) \ out=weaksblsrc

86 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-17 Figure 3.9: The spectrum of SgrB2 M on the baseline 3-4 after applied for all the bandpass corrections. Top panel is for amplitude and bottom is for phase.

87 3-18 CHAPTER 3. STRATEGIES IN CALIBRATIONS CALIBRATION WITH MULTIPLE CALIBRATORS Bandpass In some data set, there might be multiple calibrators included and any one of the calibrators might give solutions with inadequate S/N. Users might desire to use all the calibrators to achieve better S/N. smamfcal normalizes the channel visibility by the continuum vector of each record prior to solving for bandpass, which allows user to use multiple calibrators. In addition, in the polarization observations with waveplates, the bandpass solved from rr and ll might be averaged in order to gain S/N. Here is a usage of smamfcal for bandpass using multiple sources and averaging rr and ll solutions: smamfcal vis=pol690mult_rx2.usb \ edge=1,1 flux=10 select= pol(ll,rr),source(3c454.3,uranus,neptune) \ reant=3 interval=1000 weight=2 options=averrll Antenna Gains Unfortunately, neither mfcal nor smamfcal has been implemented for solving antenna gains from multiple calibrators. However, it is possible to obtain an antenna-gain table from multiple calibrators. Here is a procedure to produce a merged gain table from different calibrators based on the data taken from the observations of HII region NGC 7538 on : 1. Split the data into single-source files Assuming the multiple source uv file H30a rx0.usb.tsys.bp.bl has been calibrated for bandpass. Two gain calibrators and were interleved. Here is the observing pattern of the target source interleaved with the two calibrators: 05SEP11:05:26: SEP11:05:32:00.0 n SEP11:05:42: SEP11:05:48:31.2 n SEP11:05:59: SEP11:06:05:02.3 3c SEP11:06:15: SEP11:06:21:28.3 n SEP11:06:32: SEP11:06:37:59.4 n SEP11:06:48: SEP11:06:54:30.6 n SEP11:07:05: SEP11:07:11:01.7 n SEP11:07:21: SEP11:12:24:53.5 n SEP11:12:35: SEP11:12:41:24.7 n SEP11:12:52: SEP11:12:57:55.8 n SEP11:13:08: SEP11:13:14:27.0 n SEP11:13:25:

88 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES SEP11:13:30:58.1 n SEP11:13:41: SEP11:13:47:29.2 n SEP11:13:58: SEP11:14:04: The gain calibrators need to be splitted: uvsplit vis=050911h30a\_rx0.usb.tsys.bp.bl \ select= source( , ) \ options=nowindows mv gtest mv gtest The calibrators and are splitted into two separate uv data sets and renamed the two single-source data files as gtest and gtest. 2. Solving for antenna gains from indvidual calibrators Now, the antenna gains can be solved from the individual files using smamfcal: smamfcal vis= gtest refant=3 interval=1 \ options=nopassol smamfcal vis= gtest refant=3 interval=1 \ options=nopassol Note that the default for the input Keyword flux is used here so that the rms gain amplitude is set to 1 and no flux density scaling effect is applied to the visibilities in this process. 3. Merge the indvidual gain tables into a joint gain table Then, the disjoint gain tables can be merged into a joint one, sorting into the ascending time order, with smagpplt: smagpplt vis= gtest, gtest \ device=/xw yaxis=amp,phase \ options=gains,merge polyfit=8 \ nxy=1,6 The merged gain table is placed under gtest, replacing the old gain table. Fig.3.10 and Fig.3.11 shows the ampltidue and phase curves computed from the merged gains (red dots). For some reasons, the antenna gains derived from the two calibrators show systematic offsets. The magnitude of the offsets varies with antennas and time. smagpplt also provides orthogonal polynomial fits to the gains in an order of polyfit=8 which can be specified by users. The polynomial fits would be helpful to minimizing the systematic errors. Users might wish to use the polynomial fits instead of the original merged gains. Then, please go back to step 2 to recalculate the antenna gains for gtest and redo smagpplt with options=gains,merge,opolyfit instead of options=gains,merge: smagpplt vis= gtest, gtest \ device=/xw yaxis=amp,phase \ options=gains,merge,opolyfit polyfit=8 \ nxy=1,6

89 3-20 CHAPTER 3. STRATEGIES IN CALIBRATIONS Figure 3.10: This plot shows the gain amplitude computed from the merged antenna gains (red dots). The blue curves are the polynomial fits (8th order).

90 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-21 Figure 3.11: This plot shows the gain phase computed from the merged antenna gains (red dots). The blue curves are the polynomial fits (8th order).

91 3-22 CHAPTER 3. STRATEGIES IN CALIBRATIONS The fitted gain table is placed under gtest. The tradeoff in using the polynomial fits is that the detailed structure in the original gains might be lost. 4. Gpcopy the merged gain table back to the multiple-source file The mergered gain table in gtest is then copied back to H30a rx0.usb.tsys.bp.bl using gpcopy: gpcopy vis= gtest out=050911h30a_rx0.usb.tsys.bp.bl \ mode=copy options=nopass PHASE TRANSFER BETWEEN 230 AND 690 GHz Simultaneous observations with dual receivers (230/690 GHz) are supported at the SMA. This is a nice feature for observations at high frequency. At high frequencies of submillimeter wavelengths, strong calibrators become rare. The phase variation due to changes of water vapor in the atmosphere might be monitored by the observation of a calibrator at lower frequency (230 GHz). The 690 GHz phase appears to be linearly correlated with the phase at 230 GHz under normal circumstances. The phase variation monitored at 230 GHz might be transferred to 690 GHz. The transferred phase variation might be used to correct for phase variation in the observation at 690 GHz. This technique requires a good stability of the instruments at both 230 and 690 GHz. Based on the experiments in the spring 2005, a procedure for the phase transfer between 230 and 690 GHz is provided below for the off-line calibration concerns: Linear Regression of the phases between 230 and 690 GHz 1. Data editing and flagging Due to the poor stability of the instrument, it had been seen that phase jumps sometimes occurred on one of the receivers. If such jumps occurred often during the period of an observation, this technique would not work. It would be wise to stop proceeding to the phase transfer process. If only few bad points seen in phase from one of the receivers, please flag the data for both receivers and try the phase transfer technique following the procedure described below. 2. Linear regression of the phases between 230 and 690 GHz The observation must include a source which is strong enough at both 230 and 690 GHz so that it can be used to determine the phase correlation between 230 and 690 GHz. Observing scans of the source must be interleaved with other sources in the observing track, covering a time range as large as possible. Here we assume that the small planet Ceres was used in the observation. Please follow the detailed procedure discussed in Chapter 2 to perform the linear regression of the phases between 230 and 690 GHz. Task varfit with options=phareg gives the linear regression of the phases between 230 and 690 GHz. Bad data points due to the instrumental defects (such as phase jumps) might ruin the linear correlation. If only few bad points, then one might go back to flag the data and run selfcal again until one is satisfied with the linear correlations on each of the antennas fitted using varfit: varfit vis=ceres.221.lsb,ceres.682.lsb xaxis=time yaxis=phase \ device=/xs nxy=2,3 refant=6 options=phareg Please keep in mind that the slope table, containing the fitted slope and phase offsets for each antenna, is saved in ceres.682.lsb, which will be used in the phase transfer later.

92 3.4. SPECIAL CALIBRATION TOOLS AND PROCEDURES 3-23 Phase Transfer from 230 GHz to 690 GHz 1. Extract phase calibrator and target source In the example used for this procedure, is the phase calibrator and i1629a is the target source. The multiple source files miriad050218c 230 rx0.lsb.tsys and at 230 GHz miriad050218c 690 rx2.lsb.tsys at 690 GHz have been corrected for system temperature. Using uvsplit, one can extract both the phase calibrator and target source out of the multiple-source files: uvsplit vis=miriad050218c_230_rx0.lsb.tsys \ select= source( ,i1629a) options=nowindow uvsplit vis=miriad050218c_690_rx2.lsb.tsys \ select= source( ,i1629a) options=nowindow Rename the splitted single source files to convienent file names: mv i1629a i1629a.221.lsb.tsys mv i1629a i1629a.682.lsb.tsys mv lsb.tsys mv lsb.tsys Any bad data points must be identified and flagged using smauvplt and uvflag. The flagging must be performed for both files at 221 and 682 GHz so that identical data records are kept in both 221 and 682 GHz files. Make sure that no bad data contaminate both files. 2. Determine the antenna-based phase variation at 230 GHz Using selfcal with the phase-only option, the antenna-based gain in phase can be solved from the observations of at 221 GHz: selfcal vis= lsb.tsys interval=3 refant=6 line=chan,1,1,768 Using smagpplt, one can check up the gain solution in phase: smagpplt vis= lsb.tsys device=/xs yaxis=phase \ options=gain polyfit=5 nxy=2,3 If the gain solutions in phase at 221 GHz satisfy you (no bad points such as 90 degree phase jumps), the gain table from the 221 GHz file needs to be copied over to the 682 GHz file using gpcopy in order to create relevant header for the 682 GHz file: gpcopy vis= lsb.tsys out= lsb.tsys 3. Transfer 230 GHz phase to 690 GHz Before do the phase transfer from 230 GHz to 690 GHz, it would be wise to copy over the gain table from lsb.tsys to i1629a.221.lsb.tsys: gpcopy vis= lsb.tsys out=i1629a.221.lsb.tsys puthd in=i1629a.221.lsb.tsys/interval value=0.25

93 3-24 CHAPTER 3. STRATEGIES IN CALIBRATIONS One may make image of i1629a at 221 GHz to make sure that the selfcal solution in phase derived from at 221 GHz does work for the calibration of the i1629a data at 221 GHz. If the calibrated image of i1629a at 221 GHz did not improve its quality, please go back to find out what is wrong with the gain solutions derived from selfcal. If one is satisfied with the quality of the calibrated i1629a image at 221 GHz, one might continue on the process. Using phatrans, the phase at 221 GHz from lsb.tsys can be transferred with the antenna-based slopes and phase offsets saved in ceres.682.lsb over to lsb.tsys. phatrans vis= lsb.tsys,ceres.682.lsb, lsb.tsys \ refant=6 Using smagpplt, one may check up the transferred 690 gains in phase: smagpplt vis= lsb.tsys device=/xs yaxis=phase \ options=gain polyfit=5 nxy=2,3 If the gains in phase look normal, then one might make an image of the calibrator at 682 GHz by applying the transferred gains to the calibrator to make certain that the phase-transfer technique works for the calibrator itself. If the image of at 682 GHz verifies the phase-transfer successful, one can copy over the gain table to the target source file: gpcopy vis= lsb.tsys out=i1629a.682.lsb.tsys Using puthd, one may need to reset the interpolation tolerance interval stored in the data header: puthd in=i1629a.682.lsb.tsys/interval value=0.25 Then, one might proceed to imaging i1629a at 682 GHz using the gains transferred from the calibrator at 230 GHz. 3.5 Calibration for The Flux Density Scale After calibrations for the fringe amplitude (with T sys ), bandpass, antenna gains, and baseline-based errors, one might proceed to bootstraping the flux density scale from an observed planet to the visibility data using smaflux which relys on the SMA Planetary Models. The catalog directory of the SMA Planetary Models in MIRIAD contains two sets of data for each planet (1) a file of the Planck whole disk brightness temperature from SMA Planetary Models and (2) the planet ephemeris from JPL s HORIZONS System. Based on the disk brightness temperature and the ephemeris data, smaflux calculates the visibility structure using a uniform disk approach. In comparison the observed visibility with the model, smaflux determines the fluxdensity scale. The beta version of SMA MIRIAD contains the SMA Planetary Models. For the old version (Miriad CVS), one need to install the SMA Planetary Models separately (see Installation of SMA Planetary Models). The usage of smaflux is given here, smaflux% inp Task: smaflux vis = uvdata_rx0.usb.tsys.bp.bl.av %the uvdata set calibrated

94 3.6. HYBRID SPECTRAL RESOLUTION 3-25 select = source(ceres) mirhome = $MIR %for Tsys, bandpass, antenna %gains, baseline-based errors %select the planet observed %for flux density scale %calibration. %the home directory of one s %Miriad installation which %leads the program to find %the data directory of the %SMA planetary models. smaflux reports the scaling factor of the flux density for the uvdata set and creates/modifies the gain table. 3.6 Hybrid Spectral Resolution Note that for hybrid spectral resolutions across the multiple spectral windows, in particular when very high spectral resolutions are used for a few windows and the rest of are lower, the MIRIAD code uvaver treats all the spectral channels equally in averaging. This may cause that the noise in the high spectral resolution chunks dominates the continuum data. Users may separate the high and low spectral chunks before forming continuum data using uvaver. Or, user may resample the spectra while the data is loaded to MIRIAD using task smalod. The antenna gains might be derived from the uniformly sampled data set and then copy to the spectral line data using gpcopy. 3.7 Very High Spectral Resolution Owing to a limited number of correlator chips, a correlator mode for very high spectral resolutions involves hybrid spectral resolutions giving some empty data chunks. The task smalod has been implemented to convert SMA data to MIRIAD. In order to keep the same nomenclature for the spectral windows (chunks ) identification, each of the empty chunks are padded with a flagged chunk. The discussion on handling hybrid spectral resolution data in the previous is also relevant to handling very high spectral resolution data. A PROCEDURE OF HIGH SPECTRAL RESOLUTION DATA REDUCTION A procedure of handling very high spectral resolution data has been discussed in an SMA MIRIAD memo. The part of procedure in data loading, data inspection, corrections for system temperature should be identical to that used for handling normal resolution data. A special attention may need to take in handling bandpass and antenna gain calibrations. Assuming that rx0.lsb.tsys is the lower side band data with corrections for system temperature, the calibration part of the procedure is highlighted below: 1. Data Separation Before executing the calibration tasks, the high (chunk 16) and low (1,2,3,4,5,6,7,8,9,10,11,12) spectral data are separated into two working files, rx0.lsb.tsys.hr and rx0.lsb.tsys.lr: uvaver vis=060512_rx0.lsb.tsys select= window(1,2,3,4,5,6,7,8,9,10,11,12) \ options=nocal out=060512_rx0.lsb.tsys.lr uvaver vis=060512_rx0.lsb.tsys select= window(16) \ options=nocal out=060512_rx0.lsb.tsys.hr

95 3-26 CHAPTER 3. STRATEGIES IN CALIBRATIONS 2. Bandpass Calibration Then, we solve for the antenna-based bandpass: smamfcal vis=060512_rx0.lsb.tsys.lr select= source(neptune) edge=3,3 \ refant=3 interval=600 smamfcal vis=060512_rx0.lsb.tsys.hr select= source(neptune) \ weight=2 refant=3 interval=600 smooth=7,1,0.9 options=msmooth Note that a moving smoothing is performed prior to solving for the antenna-based bandpass in order to improve the signal for the high spectral resolution data. Please check up the bandpass solutions: smagpplt vis=060512_rx0.lsb.tsys.lr device=/xs \ yaxis=amp,phase \ options=bandpass polyfit=5 \ nxy=2,4 smagpplt vis=060512_rx0.lsb.tsys.hr device=/xs \ yaxis=amp,phase \ options=bandpass polyfit=5 \ nxy=2,4 If the solutions are satisfying, then they can be applied to the data: uvaver vis=060512_rx0.lsb.tsys.lr options=nocal \ out=060512_rx0.lsb.tsys.lr.bp uvaver vis=060512_rx0.lsb.tsys.hr options=nocal \ out=060512_rx0.lsb.tsys.hr.bp 3. Antenna Gain Calibration The phase reference source ( ) can be split from the lower-spectral-resolution multiple source file that has been applied with bandpass corrections : uvsplit vis=060512_rx0.lsb.tsys.lr.bp \ select= source( ) options=nowindow mv lsb.tsys.lr.bp % rename the file Then we solve for antenna-based gain (phase-only): selfcal vis= lsb.tsys.lr.bp \ refant=5 interval=.5 options=pha One may check up the gain solutions. If the solutions are good, then the gain table can be copied over to both the high and low spectral resolution data files: gpcopy vis= lsb.tsys.lr.bp out=060512_rx0.lsb.tsys.lr.bp \ mode=copy options=nopass gpcopy vis= lsb.tsys.lr.bp out=060512_rx0.lsb.tsys.hr.bp \ mode=copy options=nopass % reset the solution interpreation tolerance to a long time interval puthd in=060512_rx0.lsb.tsys.lr.bp/interval value=600 type=double puthd in=060512_rx0.lsb.tsys.hr.bp/interval value=600 type=double

96 3.8. EXPORT CALIBRATED DATA 3-27 One could apply the gain solutions to the data or leave the gain table in the data files and apply them while executing relevant tasks. To this point, the calibration for bandpass and antenna gains are completed. 3.8 Export Calibrated Data The calibrated SMA data can be exported to other environments for further data processing. Task fits supports conversion between MIRIAD format of SMA data and FITS format FITS to AIPS Here is an example of converting SMA data in MIRIAD format to uv-fits format: fits% inp Task: fits in = gc_rx1.lsb.tsys.cal.cont op = uvout out = /aips/fits/smacont.uvfits The converted FITS data is exported to AIPS area and can be directly loaded with AIPS Task FITLD into AIPS. This conversion also supports multiple sources. For the spectral line data, the frequency or velocity might be screwed up in a case of multi-spectral windows. Users need to pay attention to the message reported from this program and may need to reset the frequencies in other software packages. Here is an example to select a single spectral window to convert it to FITS: fits% inp Task: fits in = gc_rx1.lsb.tsys.cal op = uvout out = /aips/fits/smaline.uvfits select = window(15) % spectral window 15 SMA users are encouraged to continue the image processing in MIRIAD. MIRIAD supports a variety of features particularly for spectral line observations and mosaic imaging of a large field FITS to GILDAS FITS is quite flexible and highly standard. Because of the flexibility in headers and conventions used in uv FITS, interpreter programs in different software packages might be not good enough to understand each other. The UVFITS output from fits seems to follow the AIPS conventions. GILDAS has its own convention in the alternation between frequency and velocity. In addition, MIRIAD programs understand and distinguish between all the polarization conventions (I, Q, U, V, RR, LL, RL, LR, XX, YY, XY, YX and etc.) used in radio astronomy. This is not true for other interferometer software packages. For users who want to export SMA data from MIRIAD to GILDAS, a MIRIAD task gildas has been implemented. gildas converts Miriad uv files to FITS following the GILDAS convention for the headers of spectral line features and degenerate the polarization complex to Stokes I (OK for non-polarization projects). Here is a usage:

97 3-28 CHAPTER 3. STRATEGIES IN CALIBRATIONS gildas in=/home/data/gildas/060512uv out=060512fits.uv \ linename= SiO For users who want to degenerate complex in polarization labelling and convert a MIRIAD uv data to AIPS UVFITS, here is a usage: gildas in=/home/data/gildas/060512uv out=060512fits.uvfits \ options=aips

98 Chapter 4 Basics in Imaging, Deconvolution and Self-calibration 4.1 Introduction The sampled visibility data in a uv plane can be converted to an image with Fourier transform. The task invert is used to produce an image or a spectral image cube based on a collection of calibrated data. Even if one is only interested in continuum data, the multi-frequency imaging is still relevant. The present SMA observations can simultaneously produce 48 spectral windows with 4 GHz in each side band; and in the near future the SMA will add additional two broad bands with a total of 8 GHz per sideband. Thus one needs to handle 100 spectral windows in the imaging process simultaneously. In particular, for combining the two sidebands with a separation of 10 GHz, the total bandwidth coverage would be large and effects of the frequency smearing and a steep emission spectrum across the observing bands become significant. The frequency-dependent effects must be removed in order to make a high sensitivity continuum imaging by combining all the spectral window data. The technique of forming a single continuum image from a variety of frequencies is called multi-frequency synthesis (MFS). MIRIAD supports the MFS feature. The MFS process is useful to SMA users in imaging the continuum dust emission from distant galaxies, planets outside the solar system, young stars in the Galactic center/disk and starburst galactic nuclei. An example of making an image from raw SMA data to final image using MFS technique is shown in the sections below. In addition, we also discuss the weighting methods in the imaging process. 4.2 Weighting Weighting in the imaging process is important in order to improve the dynamic range and the fedility of a final aperture synthesized image. The ATNF s Users Guide gives a thorough discussion and description in both sampling density weighting and sensitivity weighting. The description of the weighting methods is relevant to the SMA users. For the SMA users convenience, we duplicate their discussion interleaving with some comments for the SMA specifications. Each visibility sample is given a weight in the imaging step. This weighting can be used to account for differences in the density of sampling in different parts of the u v plane, or to account for different noise variances in different samples, or to improve sensitivity to extended objects, etc. Here we briefly review some weighting schemes: 4-1

99 4-2 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION Natural weighting This gives constant weights to all visibilities (or, more strictly, inversely proportional to the noise variance of a visibility). This weighting gives optimum point-source sensitivity in an image. However the synthesised beam-shape and sidelobe levels are usually poor. Uniform weighting This gives a weight inversely proportional to the sampling density function. This form of weighting minimizes the sidelobe level. However the noise level can be a factor of 2 worse than natural weighting. Super- and sub-uniform weighting Uniform weighting computes the sampling density function on a grid that is the same size as the gridded u v plane. This results in the synthesised beam sidelobes being minimized over the same field-of-view as the region being image. Surprisingly, making the fieldof-view very large (bigger than the primary beam size) or very small (comparable to the synthesised beam) both cause uniform weighting to reduce to natural weighting. Super- and sub-uniform weighting decouple the weighting from the field size being imaged. Instead, the sidelobes in the synthesised image are minimized over some arbitrary field size, with this field being either smaller or larger than the field being imaged (for super- or sub-uniform weights respectively). Robust weighting Uniform weighting (including super- and sub-uniform weighting) minimizing sidelobes, whereas natural weighting minimizes the noise level. Robust weighting provides a compromise between the two, doing so in an optimal sense (similar to Wiener optimisation). See Dan Briggs thesis for more information (Dan Briggs dissertations). Tapering In signal processing theory, the optimum way to detect a signal of known form, which is buried in noise, is to convolve that signal with a matched filter. This filter has an impulse response which is just the reverse of the form of the signal that is being detected. Applying this principle to detecting sources in radio interferometry, the optimum weighting for detecting a Gaussian source is to weight the visibility data by a Gaussian. This is often called tapering. Using a Gaussian weight will significantly increase the detectability of an extended source. However it also degrades the resolution. Gaussian weighting can be combined with any of the above weighting schemes to achieve some form of balance between sidelobes and sensitivity. MIRIAD gives good control over the visibility weighting schemes, via three parameters and one option. fwhm controls tapering of the data. Unlike AIPS, this taper is specified in the image domain, in arcseconds. If you are interested in features of a particular angular size, then the signal-to-noise ratio in the resultant dirty image is optimised for these features if the taper FWHM is the same as the source FWHM (or source scale size). sup is used to control super- and sub-uniform weighting. The sup parameter indicates the region of the dirty beam (centred on the beam centre) where sidelobes are to be suppressed or minimized. Like the fwhm parameter, the sup parameter is given in arcseconds. The weights that invert calculates are optimal, or near optimal, in a least-squares sense to minimizing the sidelobes in the specified region of the beam. Although the sidelobe suppression region is not a direct control of resolution and signal-to-noise ratio, it does have an effect on these characteristics. Suppressing sidelobes over the entire field of the dirty beam corresponds to uniform weighting that is, we get uniform weighting if sup is set to the field size of the dirty beam; this is the default. Alternately making no attempt to suppress sidelobes (sup=0) corresponds to natural weighting. Increasing sup from 0 to the field size results initially in an improvement in resolution until the value of sup is approximately equal to the best resolution. Increasing sup beyond this results in a slow degradation in resolution. The noise level varies in a less regular way with sup. Apart from saying that sup=0 (natural weighting) gives the optimum signal-to-noise ratio, it is not possible to generalise. However the noise level will be no worse than a factor of a few from the optimum. The default value for sup is the field size (i.e. uniform weighting).

100 4.3. DIRTY IMAGE 4-3 robust: In MIRIAD, robust weighting is parameterized by a robustness parameter. Values less than about -2 correspond essentially to minimizing sidelobe levels only, whereas values greater than about +2 just minimizing noise. A value of about 0.5 gives nearly the same sensitivity as natural weighting, but with a significantly better beam. options=systemp: The basic weight of a visibility (the weight used for natural weighting, or the weight used for a point in determining the local sampling density function) is ideally 1/σ 2, where σ 2 is the noise variance of a visibility. For no Tsys information given, invert normally assumes the noise variance is inversely proportional to the integration time ( t) and the bandwidth ( ν) of a visibility. However SMA data loaded with MIRIAD task smalod contains antenna-based Tsys measurements and MIRIAD can calculate σ 2 using the formula below: [ ] Ts(i)Ts(j) σ 2 (i, j) = JyperK 2 2 ν t where JyperK = 2k/ηA and A is antenna area, η = η c η a is an efficiency, k is Boltzmans constant. The options = systemp can instruct invert to compute the basic weights using σ 2. The calculation assumes that the parameters A, η c and η a are constant across the array and the sky, i.e. JyperK is assumed to be constant. For optimizing the sensitivity of the final synthesized image, the relative weights derived from options = systemp need to be applied. 4.3 Dirty Image The MIRIAD task invert has been configured for the reduction of SMA data. The maximum number of spectral windows has been extended to a limit of 48. MIRIADtask invert forms images from visibilities. Both continuum images or spectral line cubes can be formed using invert. It can generate images or image cubes for several polarizations, as well as handling multi-frequency synthesis and mosaicing observations. MIRIADtask invert can also form complex-valued images from non-hermitian data (e.g. holography data). Appropriate point-spread functions (dirty beams) can also be generated. invert% inp Task: invert vis = sgra-star.cal % uvdata contain 24 chunkaveraged spectral points. map = sgra-star.map % name of dirt images beam = sgra-star.beam % name of dirt beams imsize = 512,512 % size of the image cell =.15 % cell size sup = 0 % weighting function; 0 for natural weighting stokes = xx options = mfs,sdb,systemp % using mfs, and produces spectral beams and normal beams; weighting by the data variance. Fig. 4.1 shows the dirty image of this data. Because of the discrete sampling function, the image is contaminated by the side lobes of the point spread function (Fig. 4.2). Apparently, a point source is dominant in the image (Fig. 4.1). The side lobes must be minimized. There are a number of algorithms that can deconvolve the dirty beam from the dirty map.

101 4-4 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION Figure 4.1: Fourier transform of the weighted visibility data or a dirty map.

102 4.3. DIRTY IMAGE 4-5 Figure 4.2: The point spread function or dirty beam of the visibility data.

103 4-6 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION 4.4 Deconvolution The point spread function depends on the observing frequency. For broad band image with multiple spectral windows, the deconvolution can utilize the multi-frequency synthesis (MFS) technique. Also, in the situation of steep spectral index (can be as large as 4 in some dust emission regions) and large frequency separation, simple convolution relationship no longer applies. A special algorithm mfclean is needed to handle multiple frequency synthesis deconvolution. MIRIADtask taskmfclean supports handling SMA data with a large number of frequency windows (chunks). In this section, a use of the traditional CLEAN algorithm by Hogbom/Clark is shown and discussed. The MIRIAD task clean performs a hybrid Hogbom/Clark Clean algorithm, which takes a dirty map and beam, and produces an output map which consists of the clean components. However, mfclean is MFS version of clean, which is useful to the SMA data with a large number of chunks. mfclean% inp Task: mfclean map = sgra-star.map beam = sgra-star.beam out = sgra-star.icmp gain = 0.1 cutoff = 0 niters = 500 region = quarter Using the dirty map and dirty beam, clean or mfclean (see above setup) can make a clean image. The clean component image is convolved with a model by using a gaussian beam (clean beam). MIRIAD task restor restores clean components to make the CLEAN map. Fig. 4.3 shows the restored clean image. Since significant residual errors remain in the uv data, the clean image appears to be distorted. We need Self-calibration to correct for the residual errors. restor% inp Task: restor model = sgra-star.icmp beam = sgra-star.beam map = sgra-star.map out = sgra-star.icln 4.5 Self-calibration and Continuum Image The technique of self-calibration can be used to make additional correction for the residual errors in antenna gains using the closure-relation technique. MIRIAD task selfcal performs self-calibration of visibility data. Either phase only or amplitude and phase calibration can be performed. Usually, it needs a few self-calibration cycles to converge the process. The following example shows the setup of self-calibration with phase only and then amplitude and phase: selfcal% inp Task: selfcal vis = sgra-star.selfcal % uv data model = sgra-star.icmp % model produced from clean

104 4.5. SELF-CALIBRATION AND CONTINUUM IMAGE 4-7 Figure 4.3: The point spread function or dirty beam of the visibility data.

105 4-8 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION Figure 4.4: Self-calibrated map: Sgr A* and the dust emission from the circum-nuclear disk at the Galactic center. interval = 10 options = phase,mfs refant = 3 % interval of solutions % phase only and mfs After the phase errors be minimized with a few step of self-calibration in phase-only, one may try selfcalibration in both amplitude and phase. selfcal% inp Task: selfcal vis = sgra-star.selfcal % uv data model = sgra-star.icmp % model produced from clean interval = 20 % interval of solutions options = amp,mfs % amp & phase and mfs refant = 3 Fig.4. shows an image with self-calibration. This image shows Sgr A* and the nearby dust components detected by the SMA at 345 GHz. 4.6 Spectral Line Image Cube The basic calibration procedure of spectral line shares the same MIRIAD tasks as discussed in the continuum data calibration. A step-by-step procedure of the reduction for SMA spectral line data (that was produced from old standard correlator mode with 24 spectral chunks per sideband) in MIRIAD was demonstrated. Here is a link to the detailed step-by-step procedure (Flow Chart and Script). In the follows, a few plots that were illustrated in the demostration are shown.

106 4.7. SYNTHESIS IMAGING WITH MULTIPLE UV DATASETS 4-9 Figs and 4.6. show the antenna-based bandpass solutions solved from 3C observed during the time between 14:00-14:45. It may be necessary to have bandpass solutions in more than one time intervals distributed evenly over the observing run so that any possible bandpass ripples can be checked and calibrated. In addition, baseline-based (and/or chunk-based) bandpass solutions with a strong point source may also be necessary for minimizing any correlator-based bandpass errors SPECTRA The spectrum of H30α on the baseline 1-2 observed from NGC 7538 is shown in Figs SPECTRAL IMAGE CUBE The antenna gains can be further calibrated using the phase reference sources ( and ) interleaved in the observation track. After applying the antenna gains to the data, it is recommended to take out the continuum emission from the data using uvlin. An image cube for the H30α line can be constructed from the continuum-free data using invert, cleaned using clean and convolved with the clean beam using restor. The final cleaned image cube of the H30α from NGC 7538 is shown in Fig Synthesis Imaging with Multiple UV Datasets MIRIAD supports imaging with multiple data sets observed in different time and different array configurations in order to improve the uv coverage. There are several ways to combine multiple data files that are discussed in this section CATENATING SUBSETS OF UV DATA In the case that the data observed in an observing run was stored into several files for some reasons, then these files are loaded into MIRIAD seperately by running smalod. The MIRIAD task uvcat is able to catenate the multiple MIRIAD files. The current version of smalod reads a MIR file. We will support loading multiple SMA data files using smalod in the near future. Here is an example to read serveral subsets of MIR files and catenate them: smalod in=/data/smadata/040722_03:47:04/ out=040722a rxif=0 refant=6 sideband=1 smalod in=/data/smadata/040722_06:34:05/ out=040722b rxif=0 refant=6 sideband=1 uvcat vis=040722a,040722b out= The output file is the catenated uvdata set with the visibility data sorted in an order of ascending time DATA ACQUIRED FROM MULTIPLE ARRAY CONFIGURATIONS For uvdata sets obtained from multiple array configuration, one needs to calibrate each array data sets separately. MIRIAD task invert accepts multiple uv data sets to produce a synthesized image. Users

107 4-10 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION Figure 4.5: The solid dots are the solutions of bandpass in ampltiude which is color-coded for chunks. The solid black curves are the 5th order polynomial fit to the solutions.

108 4.7. SYNTHESIS IMAGING WITH MULTIPLE UV DATASETS 4-11 Figure 4.6: The solid dots are the solutions of bandpass in the unwrapped phase which is color-coded for chunks. The solid black curves are the 5th order polynomial fit to the solutions.

109 4-12 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION Figure 4.7: The Spectrum of H30α in the HII region NGC 7538.

110 4.7. SYNTHESIS IMAGING WITH MULTIPLE UV DATASETS 4-13 Figure 4.8: The spectral images of the H30α line in NGC 7538.

111 4-14 CHAPTER 4. BASICS IN IMAGING, DECONVOLUTION AND SELF-CALIBRATION do not have to physically combine the data uv sets prior to imaging. The imaging routines can internally combine multiple uv data sets. For two or more uv data sets that are observed and calibrated seperately, this is a way to go. Here is an exmaple of usage: invert vis=lores.uvhanning,hires.uvhanning \ map=helixcomb_co2-1.map \ beam=helixcomb_co2-1.beam \ imsize=256,256 \ cell=0.5 \ fwhm=1.5 \ mode=fft \ line=channel,144,1146,1 Here lores.uvhanning and hires.uvhanning are the calibrated single source files obtained with the same correlator setup using a smaller array and a larger array, respectively. The linetype: line=channel,144,1146,1 allows to combine the two different array s data in a channel-by-channel basis DATA ACQUIRED WITH DIFFERENT CORRELATOR CONFIGU- RATIONS For uvdata sets obtained from different correlator configurations, the header frequency structure in the MIRIAD data could differ in the interested data files. Users might calibrate the individual data sets first. Then, using linetype in MIRIAD, users may redo the grid of the spectral visibility data along the frequency axis by selecting velocity linetype in uvaver. Once, the uvsets have the identical frequency structure, MIRIAD task invert can synthesize an imaging line cube from multiple uvdata sets. Here is an exmaple of the usage: uvaver vis=n7538.h30a_1 line=velocity,320,-220,1 out=n7538.h30a_1.vel uvaver vis=n7538.h30a_2 line=velocity,320,-220,1 out=n7538.h30a_2.vel uvaver vis=n7538.h30a_3 line=velocity,320,-220,1 out=n7538.h30a_3.vel Here n7538.h30a 1, n7538.h30a 2, n7538.h30a 3 are the calibrated single source files but the frequency structure differs from each other. The linetype: line=velocity,320,-220,1 selects the velocity domain for redoing the frequency grid, i.e. output channel number = 320, beginning channel in velocity=-220 km/s and channel width = 1 km/s. The output uv datasets n7538.h30a 1.vel, n7538.h30a 2.vel and n7538.h30a 3.vel are re-constructed to the same frequency structure and invert can make an image line cube synthesized from the three uvdata sets: invert vis=n7538.h30a_1.vel,n7538.h30a_2.vel,n7538.h30a_2.vel \ map=n7538.h30a.map beam=n7538.h30a.beam \ imsize=256,256 cell=0.2 sup=0 \ line=channel,36,100,5 The linetype: line=channel,36,100,5 set in invert is to produce 36 line image panels starting from the 100th channel with a line width averaged down to 5 km/s.

112 Chapter 5 Application Utility (UV) 5.1 Introduction There are two domains (UV and/or image) in which users can carry out analysis of the data acquired from an interferometer array. MIRIAD provides users a variety of software tools to analyze the data in either the UV domain or the image domain. In this Chapter, a few useful tasks for data analysis in the UV domain are introduced. 5.2 The JPL Catalog And Identification of Spectral Lines The MIRIAD task smauvspec supports a feature of plotting the UV spectra along with the spectral line transitions selected from the JPL catalog. This feature is useful for general MIRIAD users to identify spectral lines contained in the UV data using the JPL catalog. This method works well for a compact source at the phase center. For an extended source, the S/N of the lines may be degraded due to large phase change across the source. For weaker line in an extended source, we recommended users to carry out the line identifications in the image domain after full calibrations. Here are two exmaples to show users how to use the JPL catalog in spectral line identifications in the UV domain SCREEN INTERACTIVE Assuming that the UV data has been calibrated for bandpass and antenna gains, one might wish to average the data in a long time interval to gain a higher S/N in the UV domain. The MIRIAD smauvspec provides an option to shift the spectrum to the rest frame with respect to the source. The systematic velocity with respect to the LSR frame can be input from the Keyword:lsrvel if no systematic source velocity is included in the on-line Doppler tracking. and the observatory velocity included in the Doppler tracking should be stored in the UV data. If the source is far away from the Doppler tracking source, one must use uvredo to compute the Doppler velocities prior to using smauvspec for spectral line identification. smauvspec% inp smauvspec Task: smauvspec vis = SgrB2_M.H2CO % vis data which has been calibrated select = ant(3)(4) % select one baseline interval = 1000 % time interval for average. 5-1

113 5-2 CHAPTER 5. APPLICATION UTILITY (UV) hann = 3 % do hanning smooth options = jplcat,restfreq % to turn on the catalog reading function and shift the spectrum to the restframe. catpath = $MIRCAT/jplcat or /home/miriad/miriad_cvs/cat/jplcat % the location of the catalog directory. vsource = 61 % systematic velocity veltype = lsr % the reference frame veldef = radio % velocity definition axis = freq,both % plot frequency vs both amplitude and phase yrange = device = /xs % display device. nxy = 1,1 % 1 plot per page. log = After setup, then type go and the program will ask users to select molecules from the screen by going through all the molecular name tags supported in the JPL catalog. Here is an example to select H 2 CO for the spectral line identifications: select 1001 H-atom (<type> y to select it or n to skip it or t to terminate) n skip select 2001 D-atom (<type> y to select it or n to skip it or t to terminate) n skip select 3001 HD (<type> y to select it or n to skip it or t to terminate) n skip select 4001 H2D+ (<type> y to select it or n to skip it or t to terminate) n skip select 7001 Li-6-H (<type> y to select it or n to skip it or t to terminate)... select HC-13-O+ (<type> y to select it or n to skip it or t to terminate) n skip select DCO+ (<type> y to select it or n to skip it or t to terminate) n skip select H2CO (<type> y to select it or n to skip it or t to terminate) y yes select C-13-H2NH (<type> y to select it or n to skip it or t to terminate) t terminate selection Then, the relevant H 2 CO transitions in the frequency range will be lablled at the rest frequencies along with the spectrum plotted in a display device (Fig. 5.1.).

114 5.2. THE JPL CATALOG AND IDENTIFICATION OF SPECTRAL LINES 5-3 Figure 5.1: The spectrum of SgrB2(M) is shifted to the rest frame along with the identification of the H 2 CO transition.

115 5-4 CHAPTER 5. APPLICATION UTILITY (UV) SCRIPT AND MOLECULAR NAME TAGS The way of interactively selecting molecular species appears to be instructive but tedious. Users might like to use a script or typing a group of the selected inputs in a shell environment to run smauvspec. Here are a few steps for this approach: 1. Edit the script First, we edit the script file jpllines.csh #!/bin/csh -f #identiy spectral line using JPL catalog # smauvspec vis=sgrb2_m.h2co hann=3 strngl=-3.5 \ interval=1000 catpath=$mircat/jplcat/ \ options=jplcat,restfreq,nobase,avall \ lsrvel=61 veltype=lsr veldef=radio \ axis=freq,amp device=/xs nxy=1,1 exit The setup is similar to the previous case. Alternatively, in this case, we select more molecules and set the molecular strength strngl = 3.5 instead of the default value (strngl = 500). We require to average all the baselines and plot the amplitude in a single spectrum (options=nobase,avall and axis=freq,amp). 2. Select the molecules in the tag file Then, we mark the relevant molecules in the tag file (moltag.txt) by replacing n with y in the first column. The second column is the molecular tag number and the third column is the name of the molecules. A template of moltag.txt can be found from the web (a molecular name tag file for JPL catalog.) n h-atom n d-atom n hd n h2d+ n li-6-h n lih n li-6-d n lid n c-atom n c-13-atom n ch n ch+ n n-atom n n-atom-d-st n nh n o-atom n oh n nh3 n ch3d n nh3-v2... y h2co... y ch3oh

116 5.3. SPECTRAL LINE AND CONTINUUM EMISSION SEPARATION y hnco... y hcccn... y ocs... n br-81-o(v=1) n h2so4 n hobr-81 n cl-37-ono2 n hc7n n c7o n cloocl n cl-37-oocl n obr-79-o n se-80-o2 n c8o n obr-81-o n hc9n n c9o n hc11n We have selected five molecules (H 2 CO, CH 3 OH, HNCO, HCCCN and OCS) for identifications of the spectral lines. 3. Execute the script We now submit the script with the input from moltag.txt jpllines.csh < moltag.txt 5.3 Spectral Line and Continuum Emission Separation MIRIAD task, uvlin separates line and continuum in a spectral visibility data set. An example is given here for taking out the continuum from a calibrated uv data with single source (e.g. SgrB2 M). 1. Find interested line transition There could be many line transitions contained in the entire 2 GHz spectral window. First, one might check the spectrum using smauvspec: smauvspec vis=sgrb2_m.h2co select=ant(2)(3) \ interval=1000 axis=freq,both device=/xs nxy=1,1 Fig.5.3 shows the 24 spectral chunks conatining many line transitions. 2. Find the line free channels We now plot the spectrum on one of the baselines as function of channel numbers. smauvspec vis=sgrb2_m.h2co select=ant(2)(3) \ interval=1000 axis=chan,both device=/xs nxy=1,1

117 5-6 CHAPTER 5. APPLICATION UTILITY (UV) Figure 5.2: The spectrum of SgrB2(M) is shifted to the rest frame along with the identifications of the transitions from the molecular species H 2 CO, CH 3 OH, HNCO, HCCCN and OCS.

118 5.3. SPECTRAL LINE AND CONTINUUM EMISSION SEPARATION 5-7 Figure 5.3: The 2-GHz spectrum of SgrB2 M from the 24 chunks as function of frequency.

119 5-8 CHAPTER 5. APPLICATION UTILITY (UV) Figure 5.4: The 2-GHz spectrum of SgrB2 M as function channel numbers. From Fig.5.4, line free channels might be defined. Careful users might want to go through all the baselines to make sure that the selected line-free channels are true for all the baselines. 3. Take out the continuum The MIRIAD task uvlin since version Jan-07 can handle uv data with multiple spectral windows with options=nowin. It would be better to select line-free channels on both sides of the interested line feature so that uvlin can fit the continuum level. Also, one might choose several line-free windows in order to get a better fitting for the continuum level. Here is a usage of uvlin: uvlin vis=sgrb2_m.h2co chans=50,200,1650,1750,2300,2600 \ out=sgrb2_m.h2co.lin \ order=1 mode=line options=nowin Fig.5.5 shows the spectrum (amplitude and phase) after the continuum level is subtracted. From this point, one might go on to make spectral line images.

120 5.3. SPECTRAL LINE AND CONTINUUM EMISSION SEPARATION 5-9 Figure 5.5: The continuum-free 2-GHz spectrum of SgrB2 M is plotted from SgrB2 M.H2CO.lin in which the continuum has been taken out.

121 5-10 CHAPTER 5. APPLICATION UTILITY (UV) 5.4 Amplitude Analysis There are two tasks in MIRIAD uvamp and smauvamp allowing users to perform amplitude analysis in the UV domain. The algorithm used in uvamp differs from those used in smauvamp although both of the programs are for amplitude analysis of visibility data and/or models. uvamp performs vector average in each of the binned annular rings in a uv-plane and allows users to shift source center to the phase center. uvamp is good for a source/model with annular symmetrics such as a point source, a circular disk, a ring and etc.. Here is the usage for running uvamp to plot the amplitude versus UV distance: uvamp vis=vis_1,vis_2,vis_3,...,vis_n bin=30,4,nsec device/xs The data can be taken from multiple visibility files (visi 1,vis 2,vis 3,...,vis n). The data are divided into 30 bins along the axis of the uv distance with a bin interval of 4 nsec. smauvamp bins the data along the x-axis (time,uvdistance, uc,vc and etc.) selected by users. smauvamp does the vector average of the baseline-based visibilities in a time interval selected by users prior to performing the statistics in each bin. Both the vector averaging and bin statistics are weighted by the visibility variances. Instead of shifting the phase center, smauvamp allows users to rotate the uv plane to match the v-axis to a desired direction. smauvamp is good for sources with axis symmetrics (such as a point source, two point sources - sinusoid beating, an elongated guassian source, an elliptical disk and etc.). smauvamp also allows users to plot time variability with proper statistics in a uv-range dominated by a point source SINUSOID BEATING OF TWO POINT SOURCES Here we show an example to plot the sinusoid beating amplitude from two point sources. The following lists the parameters for the two point sources contained in the UV data: Source 1, Object type: point Flux (Jy): /- 1.78E-06 Offset Position (arcsec): Positional errors (arcsec): 1.43E E-07 Source 2, Object type: point Flux (Jy): /- 1.77E-06 Offset Position (arcsec): Positional errors (arcsec): 1.26E E-07 Here is the usage to plot the sinusoid beating amplitude as function of the uv distance: Task: smauvamp vis = /data/smadata/orion.model % uv file name axis = vc,amp % axis selection xrange = 0,130 % range in the x-axis nbin = 50 % 50 bins along the x-axis PosA = 7 % rotating the uv-plane by 7 deg to match the v axis along the line connecting the two point sources. device = /xs % plotting device The uv plane is rotated by 7 degree so that the transformed v-axis is in the direction along the connection line between the two point sources.

122 5.4. AMPLITUDE ANALYSIS 5-11 Figure 5.6: The amplitude beating for the two point sources is plotted after rotating the uv-plane by 7 degree.

123 5-12 CHAPTER 5. APPLICATION UTILITY (UV) TIME VARIABILITY IN AMPLITUDE Assuming the uv data have been calibrated for the antenna gains, one might want to plot the amplitude versus time by selecting a particular source and a uvrange in order to investigate the time variability. Here is the usage of smauvamp for this purpose: Task: smauvamp vis = mydata.uv % uv file name select = source(sour1),uvrange(25,100) % select sour1 and uvrange from 25 to 100 kilo wavelengths axis = time,amp % time for x-axis amplitude for y-axis average = 1 % averaging the visibility in time interval of 1 min for each baseline prior to doing bin statistics nbin = 50 % make 50 bins in the range of x-axis. device = /xs % plotting device 5.5 Re-Weighting Visibility After solving for antenna-based gain solutions (based on QSOs), one might realize that the gain corrections in amplitude are lower for some antennas due to some reasons (such as larger pointing errors, larger coherence loss and etc.). One might want to include those factors in the visibility weighting in the imaging process in addition to the visibility variance calculated based on the systemp temperature, bandwidth, and integration time. The MIRIAD smarewt provides a software tool to calculate the mean values of antenna gains in amplitude and multiply the visibility variances by the mean antenna gains. The re-scaled visibility variances can be used as the weights in the imaging process. Here is an example of usage: smarewt% inp Task: smarewt vis = uvdata_rx0.lsb %the input uvdata gain = uvdata_rx0.lsb %the uvdata that contains the antenna-gain table. options = report %to report the mean gains. out = uvdata_rx0.lsb.rw %the output uvdata in which the systemp (or variance) has been mupltiplied by the re-weighting factors that are calculted based on the antenna gains. Note that one must turn on the options=systemp in invert in the imaging process in order to make this re-weighting effective.

124 5.5. RE-WEIGHTING VISIBILITY 5-13 Figure 5.7: The plot of amplitude variability in time for sour1.

125 5-14 CHAPTER 5. APPLICATION UTILITY (UV)

126 Chapter 6 Application Utility (IM) 6.1 Introduction In this Chapter, a few useful tasks for data analysis in the image (IM) domain are discussed. We will give examples while the usages for the analysis of SMA images are discussed. 6.2 Primary Beam Correction for Single Field The images constructed from calibrated UV data are attenuated by the primary beam. For the SMA dish with 6 m in diameter, the primary beam of the SMA can be described by a Gaussian with a size depending on the observing frequency. MIRIAD task pbplot calculates and plots the primary beam profile for a given telescope and an observing frequency: Task: pbplot telescop = sma %telescope freq = 219 %observing frequency in GHz device = /xs %display device For a single field observation, the primary beam correction for an SMA image can be done with MIRIADtask linmos if the image header parameter Telecope=SMA: Task: linmos in = sgrb2ml.icln %the input line image cube out = sgrb2ml.icln.pbcor %the output line image cube %with correction for primary %beam Before running linmos, one may use prthd to check up the image header. If the header parameter Telescope is not properly specified, one needs to reset the FWHM of the primary beam to the image giving a type of the primary, say, Gaussian for SMA before running linmos: Task: puthd in = sgrb2ml.icln/pbtype value = gaus(50.4) 6-1

127 6-2 CHAPTER 6. APPLICATION UTILITY (IM) Alternatively, one can combine a few MIRIAD tasks to carry out the primary beam correction. First, using pbplot, one calculates the primary beam size: Task: pbplot telescop = sma %telescope freq = %observing frequency device = /xs %display device The FWHM (0.84 arcmin or 50.4 arcsec) of the primary beam at the frequency is reported on the terminal. Then, using imgen, one can generate a primary beam model: Task: imgen in = sgrb2ml.icln11 out = sgrb2ml.pbmodel factor = 0 object = gaussian spar = 1,0,0,50.4,50.4,0 This produces a primary beam model image (sgrb2ml.pbmodel). One may display the primary beam model image with cgdisp prior to correcting for primary beam. MIRIAD maths can divide the image by the primary beam model image: exp = sgrb2ml.icln11/sgrb2ml.pbmodel %the math expression for primary %beam correction. mask = sgrb2ml.pbmodel.gt.0.5 %cutoff based on model s intensity. out = sgrb2ml.pbcor %output image with PB correction. 6.3 Primary Beam Correction and Mosaicing Multiple Fields For mosaicing multiple fields, one might first read the ATNF user s Guide on Mosaicing Observing Strategies. After construction of the clean images for the individual fields, MIRIAD task linmos performs a simple linear mosaicing of input image (cubes) to produce a single output image (cube), correcting for the primary beam. Before running linmos, one needs to check up the header parameter Telescope of the images. If the header parameter Telescope is not specified, one needs to specify the primary beam type pbtype of the images (see previous section). Then, with linmos, one can perform mosaicing of the input images. Here is a usage: in = sgrb2_1,sgrb2_2 %the input clean images for the %field 1 and field 2. out = sgrb2.tpbcor %output image with bp correction. options = taper %for uniform noise distribution. The options= taper gives approximately uniform noise across the image. Fig shows a mosaicing image from the H 2 CO line observations of the field 1 (sgrb2 1 for SgrB2(M)) and the field 2 (sgrb2 2 for SgrB2(N)). 6.4 Conversions between Coordinate Systems Users often ask for converting a coordinate system of an image to others. MIRIAD task regrid supports quite a few conversions.

128 6.4. CONVERSIONS BETWEEN COORDINATE SYSTEMS 6-3 Figure 6.1: The mosaicing image of SgrB2 for one of the channels produced from the H 2 CO absorption line observations of field 1 (SgrB2(N)) and field 2 (SgrB2(M)).