next previous
Up: The synthesis telescope at


Subsections

9 Performance

9.1 Synthesized beams

The synthesized beam of the Synthesis Telescope depends on the taper applied in the u-v plane; details of synthesized beams are given in Table 6. A uniform weighting is always applied when making images from the telescope. The objectives of research with this telescope usually benefit more from better sensitivity and lower sidelobe levels than from high spatial resolution, so a Gaussian taper falling to 20% at 144L is typically applied to the u-v plane.

9.2 Field of view

The half-power beamwidth of the antennas is 107.2' at 1420 MHz and 332.1' at 408 MHz, but experience has shown that the usable field of view extends at least to the 10% level of the respective primary beams (diameters of 187' and 578', respectively). Empirical measurements show that these beams are well approximated by a function of the form $\cos^6
(q_\nu r)$. If r is the radius in degrees, q1420=30.24, and q408=9.762. Precise knowledge of the beam function is required for accurate mosaicing of images of individual fields.

9.3 Bandwidth and time-averaging effects

Wide-field imaging is an important aspect of the Synthesis Telescope, so off-axis effects must be carefully considered.

The continuum bands at 1420 MHz span 35 MHz, and, owing to differential delay effects, if the entire band were assumed to be at the nominal centre frequency there would be a reduction in point-source sensitivity of approximately 50% at a radius of 90'. This would be accompanied by a radial smearing of a point source to approximately twice the nominal beamwidth; while this preserves total flux, it is clearly unacceptable for high-fidelity imaging.

By treating the four 7.5 MHz continuum bands separately during imaging, taking into account the actual centre frequency of each, the reduction in point source sensitivity at 90' is only 5%, with a comparable amount of source distortion (see Bridle & Schwab 1989). The 3.5-MHz bandwidth at 408 MHz reduces the point-source sensitivity by about 10% at a radius of 300', again with a similar amount of source distortion.

   
Table 7: Telescope sensitivity
  1420 MHz 1420 MHz 408 MHz
  continuum spectrometer continuum
$T_{\rm S}$, K 60 60 150*
$\eta_{\rm A}$ 0.55 0.55 0.60
$\eta_{\rm c}$ 0.985 0.88 0.88
$N_{\rm b}$ 21 12 21
$N_{\rm IF}$ 2 2 1
$\Delta f$, MHz 30 $B/160^\dagger$ 3.5
W 2.0 1.33 2.0
$\Delta S_{\rm theor}$ mJy/beam 0.28 20B-0.5 3.0
$\Delta T_{\rm theor}$ K $0.071\sin\delta$ $3.5 B^{-0.5}\sin\delta$ $0.75\sin\delta$
$\Delta S_{\rm meas}$ mJy/beam 0.27 18B-0.5 3.8
 
* Includes $T_{\rm sky}=45$ K, a typical value in the Galactic plane
$^\dagger$ B is the overall spectrometer bandwidth in MHz.

Effects due to time averaging are also a factor. At the radii of 90' at 1420 MHz and 300' at 408 MHz, the visibility averaging period of 90 s reduces point-source sensitivity by about 8% in a 12-hour observation, with source distortion of similar magnitude in the azimuthal direction. The combined effect of bandwidth and time-averaging smearing is then a worst-case reduction of point-source sensitivity of about 13% at 1420 MHz and 17% at 408 MHz, with both radial and azimuthal distortions. Software written for analyzing Synthesis Telescope images takes these effects into account when determining, for example, point-source fluxes.

9.4 Noise in images

The rms noise, $\Delta S$ (W m-2 Hz-1), in an image made by a synthesis telescope is (Crane & Napier 1989)

\begin{displaymath}%
\Delta S = {{W\sqrt{2} k T_{\rm S}}\over{\eta_{\rm c} \eta_{\rm A} A \sqrt{N_{\rm b} N_{\rm IF}
{\Delta f} \tau}}}
\end{displaymath} (1)

where W is a factor that depends on the weighting scheme applied to the visibilities during imaging, k is Boltzmann's constant, $T_{\rm S}$ is the system temperature (K), $\eta_{\rm c}$ is the correlator efficiency (defined in Sect. 5), $\eta_{\rm A}$ is the aperture efficiency of the antennas, each with area A (m2), $N_{\rm b}$ is the number of baselines, $N_{\rm IF}$ is the number of IF channels, each of bandwidth $\Delta f$ (Hz), and $\tau$ is the integration time (s). $N_{\rm IF}=1$ for a single polarization, and $N_{\rm IF}=2$ when two polarizations are received.

The brightness-temperature sensitivity in K is

\begin{displaymath}%
\Delta T = {{1}\over{\Omega}} {{\Delta S \lambda^2}\over{2k}}
\end{displaymath} (2)

where $\Omega$ is the synthesized beam area and $\lambda$ is the wavelength.


  \begin{figure}
\includegraphics[width=16cm,clip]{H2136F7.PS}\end{figure} Figure 7: Images at 1420 MHz of a field containing the Hii region W5. Top left: continuum image made from calibrated visibilities after editing interference, but before application of any image processing routines. Top right: fully processed image with single-antenna data incorporated. A correction has been applied to compensate for the primary beam of the antennas. The circular boundary is at a radius of 90'. Lower left: image in Stokes parameter U. Other than editing of interference, no processing has been applied. Lower right: Hi image in one of 256 channels of the spectrometer. Continuum emission has been estimated from end-channels of the spectrometer and subtracted. No single-antenna data have been added to either of the lower two images

The factor W is unity for a naturally weighted image, but this weighting scheme is never used with this telescope. Two weighting schemes used are (a) uniform weighting, for which W=2, and (b) Gaussian tapering of uniformly weighted data, tapering to 20% at the longest spacing, for which W=1.33. The brightness temperature sensitivity is $\Delta
T=0.25\Delta S \sin\delta$ K for case (a) and $\Delta
T=0.18\Delta S \sin\delta$ K for case (b).

Table 7 shows the calculated sensitivities for each of the three telescope outputs, and compares these predictions with measured noise levels from images. At 408 MHz the thermal noise limit is not reached because of confusion.

The dynamic range achieved with the telescope varies with frequency and region observed. In the 1420 MHz continuum channel, dynamic ranges of 104 are commonly achieved. The limits to this performance are, at the low end, system noise, and, at the high end, artefacts in the vicinity of strong sources. Artefacts, which are very difficult to remove by normal processing, rise above the thermal noise limit when telescope pointing is closer than 3$^\circ$ to Cas A or Cyg A. At 408 MHz, the dynamic range achieved is about $5 \ 10^3$, for offsets greater than 5$^\circ$ from strong sources.

9.5 Imaging performance

Owing to the thorough sampling of baselines in a $12\times12$ hour observation set, the images produced by the Synthesis Telescope have very low sidelobe levels and excellent sensitivity to extended structure. This can be seen clearly in Fig. 7, which shows part of the data from one 1420-MHz field from the Canadian Galactic Plane Survey. The extended object which dominates Stokes I images of this field is the Hii region W5 (IC 1848), in the Perseus arm of the Galaxy at a distance of $\sim2$ kpc. The "raw'' image (top left), is made from the calibrated visibilities after editing interference, but before application of any image processing routines. The large emission regions are surrounded by depressions, produced by the lack of data for baselines < 3L. The final image, after incorporation of single-antenna data, is shown at upper right.

The raw Stokes U image (lower left) shows structure on many scales, totally unrelated to that seen in the I image. These polarization structures arise from Faraday rotation along the line of sight - they are not intrinsic features of the emission source. The large elliptical polarization structure is discussed briefly in Sect. 10.

The lower right panel shows one of the 256 spectral-line channels; the data have been edited and a continuum image, formed from spectrometer channels free of Hi emission, has been subtracted, but no single-antenna data have been added.


next previous
Up: The synthesis telescope at

Copyright The European Southern Observatory (ESO)