Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 339, 897-903 (1998)

Previous Section Next Section Title Page Table of Contents

2. Observations

2.1. May 1995 - VLBA test observations

The observations in May 1995 were designated a test of 43 GHz phase-referencing techniques. For target masers we selected the brightest detections of Sjouwerman et al. (1998) lying close to Sgr A*. We used all 10 antennas of the VLBA and attempted to detect three strong SiO masers. The observational setup is described in Table 1, the antennas used are listed in Table 2. The source cycle times (the time on the reference calibrator plus the time on the maser target) was varied between 30 and 240 seconds during the run. The cycle time should be a trade-off between the atmospheric phase stability and signal to noise ratio on our [FORMULA] 2 Jy, extended phase-reference source Sgr A*. For unknown reasons the Pie Town (PT) antenna failed, and bad weather affected much of the other data on the crucial short baselines of the array. Because of considerable scattering at the GC, the longer baselines were of little use for the detection of the scatter broadened masers and Sgr A*. No masers were found, but we were able to detect our intended phase-reference source Sgr A* with sufficient signal to noise for fringe-fitting in 15 second scans on the shorter baselines. For longer cycle times ([FORMULA] 50 seconds, Fig. 1) phase coherence was either lost, or phase referencing would leave phase ambiguities unresolved.


Table 1. Observational summary


Table 2. Array and station setup

[FIGURE] Fig. 1. RMS phase error [FORMULA] in May 1995 for different time intervals [FORMULA]. Different symbols identify different baselines. For [FORMULA] 50 seconds the proper number of cycles is more difficult to recover, resulting in an unreliable calculation of the structure function. From the plot we read a coherence time [FORMULA] of about 20 seconds for the 'good' baselines on J1733-130

To estimate the coherence time in May 1995, we plotted the temporal phase structure functions (Fig. 1). For a particular baseline and time interval [FORMULA], the phase difference [FORMULA] of the visibility phases after antenna calibration was measured (see top frame of Fig. 2). The coherence time is taken as the largest time [FORMULA] for which the average RMS phase error [FORMULA] is less than 1 radian. To calculate [FORMULA] only short baselines on our calibrator source J1733-130 were used; Sgr A* is too weak. We find a coherence time of approximately 20 seconds for 43 GHz in the May 95 observations. We decided to use the shortest possible cycle time of 40 seconds, and to use only the inner 6 VLBA antennas in further observations.

[FIGURE] Fig. 2. Phases on a short baseline (FD-KP) for Sgr A* during a part of the 43 GHz test (May 95). The different cycle times can be seen by the source gaps: for the first [FORMULA] seconds the interferometer was continuously tracking Sgr A*; for the next [FORMULA] seconds a cycle of 15 seconds on - 15 seconds off Sgr A* was employed; during the third period of [FORMULA] seconds we used a cycle time of 30 seconds on - 30 seconds off Sgr A*; finally in the last period a cycle time of 30 seconds on - 90 seconds off Sgr A* was used. In the top frame we display the phases after only a-priori calibration, i.e. corrected for amplitude and a constant delay. In addition to atmospheric effects a large scatter is also caused by the low signal-to-noise ratio on Sgr A*. The middle frame displays the phases for Sgr A*, after fringe-fitting and interpolation. With a third order polynomial fit to the phases we have taken out most of the residual phase slopes and ambiguities (the bottom frame)

2.2. VLBA and phased VLA observations - December 1995, January 1996

The second series of experiments took place in December 1995 and January 1996, and included the phased VLA (in B-array and in transition from B- to CnB-array, respectively). The VLA proved to be crucial for this experiment, providing a) short baselines to the VLBA antennas, b) better sensitivity, and c) a contemporary check on the flux densities of the maser targets (see Sect. 3.2). In addition we used a total bit rate of 256 Msamples/s for the second epoch, increasing the sensitivity by oversampling.

The VLBA antennas are designed for fast source switching, the VLA is not 1. Actually, it is the VLA correlator software that requires 20 seconds in between source changes and makes phase-referencing with a single VLA antenna, or with the complete VLA as one phased array for our project impractical. To circumvent this problem, the VLA was divided into two sub-arrays. One sub-array was used to observe the phase-reference source (Sgr A*), the other the maser target. The antennas were carefully divided in two interspersed sub-arrays, such that the atmospheric effects, and thus the calibration, would be comparable for each of the sub-arrays (Weiler et al. 1974). This implies that only one of the circular polarizations could be recorded for the phase-reference source, and the complementary polarization for the masers. For sensitivity reasons, we chose 5 antennas in the phase-reference array, and 8 in the maser/target one.

Every 70 minutes, all antennas were directed to J1733-130 for pointing scans and calibration. Thereafter, while the VLBA was independently performing its phase-reference schedule with a (20 + 20) second cycle time, both VLA sub-arrays were observing in simultaneous blocks of 260 seconds: 60 seconds on Sgr A* to phase up the sub-array and 200 seconds continuously on either Sgr A* (sub-array VLA-Ref) or a selected maser source (sub-array VLA-Mas). This complex schedule was implemented by creating a VLBA schedule that also drove the recorder at the VLA. However, the frequency and pointing setup of both VLA sub-arrays were created by hand. A summary of the observational setup is given in Table 1; the array setups can be found in Table 2.

2.3. Data reduction

After the path of data editing, bandpass correction and calibration of antenna gains, a constant delay based on fringe finder observations was taken out. The resulting phases, shown in the top frame of Fig. 2, contain effects of the atmosphere, source structure and possibly the effects of an improper correlator model, including positional errors for the sources and telescopes. However, we expect that the largest effect will be atmospheric, in particular tropospheric. At 43 GHz, the absolute flux density calibration is determined with a relatively high uncertainty of about 30%. For calibration of the VLA gains, note that with a 40 second cycle time, it is essential to specifically ask for 5 to 10 second calibration entries in the log files of the VLA sub-arrays (in the "cal" files) as the default of 30 or 60 seconds is too large to calibrate the data. For the Jan 96 experiment we approximated the missing VLA sub-array system temperature by scaling Tant/Tsys with the single antenna value system temperature.

The implementation of phase-referencing with respect to Sgr A* is simply done by fringe-fitting on Sgr A*, and applying the solutions for phase, delay and rate to the maser data. To the extent that they are identical in the direction of Sgr A* and the target, all aforementioned effects can be calibrated. However, a linear phase connection using rate solutions for Sgr A* did not remove all phase ambiguities in the maser data (Fig. 2, middle frame). The bottom frame in Fig. 2 shows the slight improvement with a third order polynomial fit to the phases. The latter fit yielded the best estimates for the maser source phases (relative to Sgr A*). Following the calibration we may attempt to detect the masers without further (self-)calibration.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1998

Online publication: October 22, 1998