The GMRT visibility database is recorded in the LTA format. The database is always multi-channel and is composed of zero or more scans. Each scan represents the data for one pointing direction and is composed of zero or more data records. The database begins with an ASCII header referred to as the global header which carries information about the observations common to all the scans. Each scan begins with an ASCII header of its own, referred to as the scan header. The scan header constitutes information relevant for the scan alone (e.g. source RA, DEC, object name, information about the frequency contents of the data, time reference information, etc.).
The LTA data manipulation libraries are organized in three layers. The lowest layers is written in C and is responsible for parsing and reading/writing the LTA database headers. It is also responsible for maintaining the consistency of the output LTA database. The interface for reading from and writing to the global and scan headers in the memory is provided by this layer. Higher level C and FORTRAN bindings are provided via light weight wrappers, which hide the details of managing the linked lists etc. However, using these interfaces, only one LTA database can be opened in a single application. Using the C++ interface, a single application program can handle any number of databases. This interface also provides many more astronomically useful features while manipulating the visibility database, and is the recommended interface to use. These libraries maintain two copies of the global header in the memory as a doubly linked list for the entire lifetime of the application (unless explicitly destroyed). Hence all changes, if any, in the output are represented in the output-header in the memory, which becomes the header of the output LTA database. One of them represents the input database while the other applies to the output database. However, only the latest scan header is maintained in the memory.
At even higher level, the C++ classes (also referred to as C++ objects) called LTAFMT, LTAVIEW and LTARec encapsulate the LTA format. Any number of these objects can be instantiated in a single C++ application allowing the handling of multiple LTA databases. These objects, briefly described below, were used in application programs for data calibration and testing some automatic data filtering algorithms. There are other C++ objects which were used for maintaining and manipulating antenna gains and other tables (in the memory and on disk), which are not described here. Those are described in a set of NCRA Technical reports.
The LTAFMT object is the base class for handling LTA database. This object encapsulates the functionality which is expected to be a common factor between all classes which manipulate LTA databases and perform higher level operations (e.g. see the LTAVIEW object below). It provides the I/O interface to the LTA headers, and presents the visibility data as a two dimensional array of complex numbers. The data-records, along with the visibility data, also carry other auxiliary data (e.g., parameters for fringe stopping, fringe rotation, FSTC, etc.). The interface to this auxiliary data is also provided by this object.
All higher level objects derived from this class (also referred to as derived classes) would have all the functionality of LTAFMT object plus the additional functionality of the inherited class. Almost all the methods of the LTAFMT (functionally, the services offered by this object), have been declared as virtual which allows the derived objects to modify the behavior of these methods. A redefinition of these methods in a derived object would override their functionality. This leaves the scope for development of higher level objects with minimal effort.
This is an object derived from LTAFMT and is designed to provide a view of the data with selection on the source name and integration in time and/or frequency applied to the data.
With a source selection, this object behaves exactly as LTAFMT except that it will give a view of the database with scans only of the selected source. The source selection can be specified via a POSIX egrep compliant regular expression (with the exception that '+' regular expression operator has been disabled since its interpretation as a regular expression operator causes conflict with the astronomical source naming convention where '+' appears as a valid character in the name). Data selection based on scan number is also possible.
Vector, scalar and robust averaging algorithms are built into this object and can be used to average the data in time or frequency. The averaging interval is supplied as the number of seconds of effective integration required. LTAVIEW will then average the data for the closest possible interval of time. This averaging of the data is, however, restricted to the scan and does not cross the scan boundary. Most of the application programs use this object for data manipulation.
Data recorded in the LTA database is composed of the data timestamp and the auxiliary data, followed by the real and imaginary parts of the visibility and self correlation data for each baseline for all recorded frequency channels. The LTAVIEW and LTAFMT objects provide access to the raw contiguous buffer of floating point numbers in the memory representing the complex visibility data record.
The LTARec object provides an astronomically useful view of this raw buffer. It is initialized using an LTA database object (LTAFMT or LTAVIEW). It can then be used to access the data timestamp, auxiliary data indexed by the antenna number, and the visibility data as a two dimensional array of complex numbers. The array indexing operator (the 'operator[]' in C++ syntax) is overloaded in this object to allow indexing inside the array of visibility with the baseline and channel number as the index. It also provides services for mapping between the baseline number and the two antenna numbers which constitute the baseline.