Analysing a few pointings in AIPS
Several of our colleagues at NCRA have enquired about the comparison between the results generated by the data analysis pipeline as compared to those generated manually. While these have been compared for long observing runs and the images are found to correlate well, we have also now compared three pointings from the TGSS analysed in AIPS and the AIPS++ pipeline. This has helped appreciate the effort required to analyse each of the snap-shot fields interactively and quantify the person-hours that are required to generate these images.

We analysed data of three fields (each with snapshots at five hour angles) in NRAO AIPS and compared these with the images generated by the AIPS++ pipeline. We compared the images in terms of (1) noise statistics - we note that in all the three fields, the rms noise on the image generated using the AIPS++ pipeline was better by a factor of 1.5 as compared to the AIPS image. Morever, the higher rms noise in the vicinity of the strong emission region in the south-east of the image is confined to a smaller region in the image generated by the pipeline as compared to the image generated interactively in AIPS. This was also commonly noticed. (2) sensitivity to extended source structure - in the figures given on the right the topmost figure is at 150 MHz generated by the AIPS ++ data analysis pipeline, the middle one is at 150 MHz generated using an interactive session in AIPS and the bottom one is the 1.4 GHz NVSS image. Notice the extended jet-like structure connecting two emission regions in the source (located north-west of the centre) is detected in the 150 MHz image generated by the pipeline and in the NVSS image. The image generated in the interactive session in AIPS still shows artifacts at the level of the bridge emission and hence has missed detecting it. (3) flux density of the sources - the flux densities of the sources in both the 150 MHz images are comparable.

In the interactive data analysis session in AIPS, a single channel of the amplitude calibrator was gain calibrated after removing bad data. Bandpass gain tables on the same calibrator were then obtained and applied to the target field data. The wide-band data were then averaged (appropriately to avoid bandwidth smearing), phase calibrated, imaged and iterated through 3 rounds of wide-field phase self-calibration and one last round of phase+amplitude self-calibration. Due to the large primary beam at 150 MHz, several facets mapping the primary beam were defined, imaged and deconvolved. Even on a fast computer, this was time consuming. Bad data were identified and removed at all stages. The image shown in the adjoining figure is a result of this analysis. The manually generated image does not detect the faint jet-like feature and also shows widespread artifacts around the strong source, primarily due to (1) remnant corrupt data in the interactive procedure and (2) uncorrected phase errors. Since the AIPS++ pipeline has been optimised to search for and excise bad data based on its statistics, it is more robust and removes most of the corrupt data. Moreover, it runs through seven cycles of phase self calibration, removing bad data at each stage after the third round of phase self-calibration thus leading to better image noise statistics and sensitivity to low surface brightness features. We note that the analysis in AIPS took a total of about 5 person-days to obtain the self-calibrated maps of three fields observed on the same day. Since the data was from the same day, the same amplitude, bandpass and phase calibration tables were used. This is enormous when compared with a few person-hours that are sufficient when the data is analysed using the AIPS++ pipeline. Moreover the quality of the images generated by the AIPS++ pipeline which has been optimised to calibrate the data and remove bad data are superior and it uses computer-hours as compared to person-hours. We present the maps obtained in AIPS++ and AIPS here and also compare the results with the NVSS at 1.4 GHz and VLSS at 74 MHz.

Some figures showing the images from the analysis pipeline, interactive session in aips, NVSS and VLSS are available here. Contour plots showing extended source morphology seen by TGSS at 150 MHz and the NVSS at 1.4 GHz are shown here.

More results will be displayed in coming days.