2. Observations and reduction
The observations were done on February 2nd 1995 at the 3.6m telescope at La Silla, Chile, with the Thermal Infrared MultiMode Instrument (TIMMI) (Käufl et al. 1994). Ground-based imaging in the thermal infrared is normally done by chopping the secondary and/or nodding the telescope. TIMMI uses both techniques in such a way that the nodding is done by exactly the chopper throw in order to maximize the efficiency. The detector used was a 64 64 pixel Ga:Si array, with a cut-off wavelength of 17.8 m. HR Car was imaged in the broad-band N filter ( =10.10 m and =5.10 m) and in the narrow-band [Ne II ] 12.8 m filter ( m and m). The log of the observations is shown in table 1. The seeing was estimated to be 1 0, using the standard star Cen. For flux calibration we used Cen and adopted 164 Jy at 10.1 m and .
Table 1. Log of the TIMMI observations.
2.1. data reduction
The final N-band image is the result of averaging 29 separate images, taken at two different pointings in order to remove effects that could be due to the position of the image on the array. (E.g. c100143 is the average of 22 separate images, each being the difference between an on-source image and an off-source image.) The [Ne II ] image is the result of 22 separate images. Each image is the average of 20 nodded exposures (10 positive minus 10 negative ones), each of which were cleaned of bad pixels and bad pixel rows. The distribution of bad rows was not random but certain rows (30,33,52 and 61) were of a lesser quality in many of the images. These occasional bad rows are mainly due to the fact that for some readouts there is an erratic but small offset to one of the ADCs. In the case of the final [Ne II ] image this means that the area 3 arcseconds North of the central star (rows 30 to 33) is less well determined than the Southern area. A single flat-field for both images was created by fitting a second order polynomial to the measured brightness of the standard star CMa at several positions on the array. This removes the global structure of the array, but not the pixel-to-pixel variations. Effectively these are removed by subtracting negative images from positive ones. Only the images that were made with the f=60 lens were used to compute the average image. For these 1 pixel corresponds to 0 336. The images with the f=40 lens have a larger field of view; all the main features seen in the f=60 images were also seen but no emission beyond the field of the f=60 images was detected.
Both the N-band and the [Ne II ] images were cleaned of high frequency noise using a low pass filter. Especially for the [Ne II ] image this greatly enhances the quality of the image. The spatial frequency of the photon noise of one per pixel is much higher than the spatial frequency of the "real" signal which mainly depends on the PSF, which is of the order of 3 pixels FWHM. Therefore a clear distinction can be made and the photon noise can be filtered effectively from the total signal. In applying the low pass filter we made sure that the FWHM of the beam did not increase.
To test the reality of the features found in the N-band image we treated the c086099 image and the offset image c100143 separately and compared those. This showed that all the features in the images that we present are seen in both images and are therefore not due to detector effects or to an over-interpretation of residual noise.
© European Southern Observatory (ESO) 1997
Online publication: June 30, 1998