Observational procedure

For $T^a \ll T^{sys}$, the amplitude of the visibility function $V$ is proportional to the flux density of the source and the normalized visibility amplitude is converted to flux density units using the measurement of $V$ for a source of known but non variable flux density (referred to as the flux density calibrator). Also, the complex antenna based gain can potentially vary as a function of time. These slow time variations can be corrected using periodic observation of a source of known structure, usually an unresolved source, referred to as the phase calibrator. The flux density of the phase calibrators, the flux density of which is potentially variable over time scales of days, is also calibrated using the flux density calibrator. Each observation therefore requires at least one observation of a flux density calibrator and periodic observation of phase calibrators to properly calibrate the data.

The VLA flux density calibrators, 3C48 and 3C286 was used for all observations. One of three calibrators close to the Galactic plane namely, 1830-36, 1709-299 and 1822-096, were used as the phase calibrator. A flux calibrator was observed for $\sim10$ min at the beginning and at the end of each observing session and the phase calibrator was observed for $\sim10$ min at an intervals of $\sim30$ min.

The planned periodic injection of noise from a calibrated noise source to measure the system temperature has not yet been implemented for the GMRT. As a result, the system temperatures for the flux density calibrator fields ( $T^{sys}_{Cal}$) and the field being mapped ( $T^{sys}_{s}$) must be measured and a correction equal to $T^{sys}_s/T^{sys}_{Cal}$ be applied as part of the flux density calibration. $T^{sys}_{Cal}$ was measured to be $\sim 110$ K while $T^{sys}_s$ was estimated from the all-sky maps at 408 MHz (Haslam et al.1982; Haslam et al.1995; Haslam et al.1981). For a few fields, $T^{sys}_s$ was also measured at few points around the source of interest and the measured system temperature was consistent with that estimated from the 408 MHz data to within $\sim 10\%$. None of the phase calibrators used for these observations are known to be variable over few hours. These calibrators were therefore also used as secondary amplitude calibrators to effectively correct for any slow variation in the receiver temperature.

The observing schedule if the GMRT on-line array control system can be supplied via a computer readable file. This file, apart from a few other system-related commands, contain instructions about the source to be tracked and the integration time on each source. However since the feedback of the antenna pointing status is not used, this file needs to be tweaked by inserting delays between various commands. For example, the amount of time taken for antennas to move from one tracking direction to another varies from antenna to antenna, and appropriate time delays must be inserted between the various commands in this file to make sure that the data recording begins only after all antennas have reached a given tracking position. The commands for alternately tracking the phase calibrator and the target source were put in an infinite loop, which allowed the above mentioned periodic observations of the phase calibrator. However, since only an infinite loop is possible, the observations had to be manually terminated at the end of observing session. Information about the status of the antennas as well as a mechanism to derive time-of-the-day information in the command syntax used in this file is highly desirable and will allow a better and more automated observing session.

On-line monitoring

The data could be corrupted due to a number of reasons including (1) RFI, (2) catastrophic hardware failures (irrecoverable in a short time), (3) intermittent hardware failures from which a quick recovery is possible (e.g. breakdown of communication between correlator control hardware and software), (4) antenna-based breakdown (for e.g., failure of the servo system resulting in stoppage of source tracking), (5) failure of the communication link between the antenna based computer (ABC) and the on-line control computer at the CEB, (6) power supply failure to some of the antennas, (7) loss of phase-lock for the four local oscillators (LOs) or a phase jump in the LOs, (8) onset of ionospheric scintillations, (9) problems in the online array control, (10) problems in the correlator control software/hardware, etc. Data affected by any or many of these sources manifests itself in various forms in the recorded data. Such data needs to be flagged from the database before it is used for mapping. The flagging information for the database was generated by on-line monitoring of the critical array control parameters as well by off-line examination of the data itself. The procedure used for this is described below.

System monitoring

It was noticed that problems related to items (3), (4), (5) and (10) listed above, occurred frequently enough to require careful monitoring of various related telescope parameters as well as on-line monitoring of the visibility data. The GMRT on-line array control system maintains a large amount of information about the status of the various sub-systems. This information is updated once every few seconds and is available in the shared memory resource of the control computer. Any arbitrary information from this resource can be extracted using the table^7.1 program. This program was used to extract the following information as a function of time:

• the packet sequence number

• the azimuth and elevation angles of the antennas

• the error between the target and current antenna azimuth and elevation angles

• the antenna servo system related flags

The table program produces the output in the form of a table which was supplied to a shell script which generated an alarm if any of the following conditions occurred:

• the error on azimuth or elevation angles of any of the antennas exceeded a threshold value. This usually was an indication of the antenna not receiving the the periodic tracking instructions from the GMRT on-line array control system computer

• the packet sequence number did not change for a threshold length of time. This usually indicated loss of communication between the antennas and the GMRT on-line array control system computer

• the flag for servo brakes was raised for any of the antennas. This usually indicated either a malfunctioning antenna servo system and/or the detection of extra antenna load (due to wind loading or due to some spurious signal indicating extra load)

In case of any of the above problems, manual intervention was required. However, this rather primitive ``automation'' did help enormously in long observing sessions. This procedure, while already quite useful, should ultimately be made part of the link between the GMRT on-line array control system and the correlator control software to (1) record on-line flagging information, and (2) control the recording of the data.

Data monitoring

On-line monitoring of the visibility data was done in two ways. First, the matmon^7.2program was used to monitor the normalized correlation coefficient for the phase calibrators. This program displays single integration cycle snapshots of the amplitude (or the phase) of the visibility data for all baselines in the form of a matrix. This display was used to determine the general health of the system before starting the observations and was useful in quickly locating catastrophic problems. Once the observations were started, the data from the correlator was monitored using the programs xtract, rantsol and closure (see Chapter 3). The amplitude and phase from all baselines were continuously displayed as a set of stacked scrolling line plots using the program oddix (which uses xtract and the plotting package of the GMRT off-line software). This display provided a $1-2$ hour long snapshot of the data covering two or more observations of the phase calibrator. Since most problems in the system can be detected using the data from the phase calibrators, the onset of a problem in the data between two calibrator scans was easily detected using these plots.

The output of rantsol (namely, the antenna based complex gains) for the calibrator scans was also similarly plotted. These plots provided information about the health of the system containing all the data on the calibrator scans. Problems ranging from a significant loss in antenna sensitivity (e.g. due to change in the antenna pointing offset across $HA=0$h seen for some of the antennas) to the onset of closure errors due to a malfunctioning correlator or the onset of ionospheric scintillations were quickly detected from these plots.

The output of closure (the closure phase for all possible triangles) was supplied to another program which raised an alarm if the closure phases exceeded a threshold value for a threshold length of time. This procedure quickly identified time varying correlator related problems quite effectively.

The output of the table and closure programs as a function of time were also saved in a file. This data was later examined and used to generate a flagging table readable by AIPS. These procedures, effectively generated first order on-line data flagging information, which is the crucial first step in improving the final data quality. Lack of time did not permit implementing these procedures as part of the on-line array control and correlator control software, but must be done in the near future to improve the reliability and quality of the final data.