2.3 VLA observations and data reduction

The observations (Table 1) were made using all four configurations of the VLA at centre frequencies of 8.46 or 8.44 GHz and a bandwidth of 100 MHz (the slight difference in frequencies between configurations has no measurable effect). They were reduced in the AIPS software package using standard self-calibration and imaging methods, with one major exception, viz. an iterative technique used to combine data from the different VLA configurations when the compact core had varied significantly between observations. In order to make the best possible images, we needed to adjust for these changes in the compact core, as well as for slight inconsistencies in the amplitude calibration of the four observing runs.

We therefore adopted the following procedure:

1. Image and CLEAN the data from the widest (A) configuration, ensuring that the compact core is centred on a map pixel.
2. Self-calibrate, initially adjusting only the phases, then the amplitudes, until the best image is obtained.
3. Use the CLEAN components from this image as a model for phase-only self-calibration of the next widest (B) configuration.
4. Image both datasets at the same resolution in order to measure the flux density of the core.
5. Adjust the (u,v) data for the larger configuration by adding or subtracting the appropriate point component to equalize the core flux densities.
6. Concatenate the two (u,v) datasets, image, CLEAN and self-calibrate phases and amplitudes as in steps (i) and (ii).
7. Split the datasets apart again and check that the core flux densities are consistent. If not, repeat steps (v) and (vi).
8. Add further VLA configurations using steps (iii) to (vii).

Two sets of images in Stokes I, Q and U were made from the combined four-configuration data set, one with full resolution (Gaussian FWHM 0.25 arcsec) and the other tapered to give a FWHM of 0.75 arcsec (Table 2). Both the Maximum Entropy and CLEAN algorithms were used to compute deconvolved $I$ images. The compact core was subtracted from the data before Maximum Entropy deconvolution, and added in again afterwards. All images were restored with the same truncated Gaussian beam.

The result of differencing the Maximum Entropy and CLEAN I  images was a high-frequency, quasi-sinusoidal ripple of near zero mean whose amplitude increased with surface brightness. This artefact clearly originated in the CLEAN image and such ripples are indeed known to be characteristic of instabilities in the CLEAN algorithm (Cornwell 1983). There was no evidence for any differences between the two images on larger scales. We therefore use only the Maximum-Entropy I images in what follows. The Q and U images were CLEANed. A first-order correction for Ricean bias (Wardle & Kronberg 1974) was made when deriving images of polarized intensity. Apparent magnetic field directions were derived from images of Q and U corrected for Faraday rotation using results from a six-frequency analysis of the polarimetry of this region at a resolution of 1.5 arcsec FWHM to be published elsewhere. The maximum correction is $\approx$9° and Faraday depolarization is negligible.

The resulting images are almost noise-limited (Table 2), and the excellent (u,v) coverage and signal-to-noise ratio allow a good representation of jet structures on a wide range of scales.

Table 1: Journal of observations

Configuration	Frequency MHz	Date	Integration time (min)
A	8460	1996 Nov 12	606
B	8440	1994 Jun 6, 14	818
C	8440	1994 Dec 4	242
D	8440	1995 Apr 28	69

Table 2: Images and rms noise levels.

Resolution (arcsec)	rms noise level ($\mu$Jy / beam area)
	I	Q/U
0.25	5.5	6.1
0.75	6.9	5.5

2002-06-13