Over 400 hours of observations were conducted on the Karl Jansky Very Large Array in New Mexico. The configuration of the array resulted in the survey being sensitive to sources with angular sizes ranging from 1.5'' to approximately 15''.
Full details on the parameters of the survey are presented in in the CORNISH design paper: Hoare et al. 2012
The Survey Area
The target area covers the whole of the northern GLIMPSE region of the Galactic plane: 10o < l < 65o and |b| < 1.1o. The false-colour image below shows the area as seen by the MSX satelite in the mid-infrared waveband.
Pointing Pattern and Sensitivity
The 110 square-degree target area is divided into 42 observing blocks, each of which contains 180 - 220 fields arranged in a hexagonal grid. This grid is a scaled version of the hexagonal NVSS pointing pattern (NRAO VLA Sky Survey, Condon et al. 1998), with a separation of 7.1' between adjacent pointing centres, compared to the 8.5' full-width half-maximum of the primary beam. During the reduction process individual fields were imaged out to an 8' radius, resulting in a sensitivity uniform to 10 percent when combined into a mosaicked image.
Each field was observed as two 50 second 'snapshots' separated by 4 hours, maximising the uv-coverage and minimising the elongation of the synthesised beam. Fields at declinations greater than -15o were observed using the VLA's B configuration, while fields at lower declinations were observed using the BnA configuration, which compensates for beam distortion at low elevations. The final root-mean-square noise level is lower than 0.4 mJy/beam, however, where very bright sources are present this can be at twice as high with significant artifacts in the field.
Observations were conducted in two seasons: the inner 25o were observed from July - September 2006, while remaining 30o were observed from September 2007 - February 2008.
Frequency Setup and Array Configuration
To image the full 8.5' primary beam these observations were carried out in pseudo-spectral line mode. This maximised the sensitivity and allowed the widest possible field of view. The two 25 MHz bandwidth spectral windows were tuned to adjoining frequency bands, each of which was sampled by eight 3.1 MHz channels, degrading the peak response by only a few percent at the edge of the primary beam. Due to hardware limitations, only the stokes RR and LL polarisations were recorded meaning that linear polarisation information was not available.
The data was reduced using a combination of the NRAO AIPS software and the experimental Obit software package developed by Bill Cotton. Calibration was performed using standard AIPS tasks, while the Obit routines IMAGER and MOSAICUTIL were used to image and mosaic the fields, respectively. IMAGER implements an algorithm to detect and automatically clean emission in the fields (see this NRAO memo) and also searches the NVSS catalogue for strong radio sources outside of the primary beam. Bright sources in the sidelobes may project artifacts into wide-field images and these regions were imaged and cleaned in parallel with the primary beam area.
The basic data product of CORNISH is a square mosaicked image tile, measuring 20' on a side and containing data from up to 30 fields. Pipeline-reduced image tiles are now available to download. A cutout server provides access to custom images for user-specified positions within the target area. In adition, the calibrated uv-data is also available as a data-product via a position query interface.
Likely the most useful data product is the CORNISH high-reliability catalogue, from which several categories of identified sources (UCHII regions, PN, etc.) may be extracted.