Efficient mapping of faint sources with a multichannel bolometer: A ``zero order'' sky--noise suppression.

Observations: We have observed NGC2146 using the 7--channel MPIfR bolometer on the 30m telescope in the ON-OFF mode. Since the individual beams (pixels) of the array have a spacing of about 21 arc sec, the array is displaced by 4 to 5 arc sec between individual ON-OFF measurements in such a way that we obtained a fully sampled map. With each map (coverage) consisting of 20 ON--OFF measurements an area on the sky of about 1 arc minute diameter is mapped. This observing procedure assumes that the `OFF--position' is emission free, thus we took a fairly large separation (chopper throw) of 120 arc sec between the ON and the OFF beam. In order to obtain a better S/N, we mapped the galaxy 8 times, displacing individual coverages in order to map a slightly larger field.

Data Reduction: After calibrating the data, they are regridded into RA/DEC. A contour plot of the original data is shown in the upper part of Fig. 3. The data is heavily affected by sky--noise. As mentioned before, because of the size of the galaxy, we had to use a large chopper throw of 120 arc sec and this increases the sky--noise in particular under the average to poor weather conditions as occurred during our observations.

By inspection of the individual ON--OFF measurements we found that all seven channels often showed a large offset either positive or negative, suggesting that this might be a major component of the sky--noise. Hans--Peter Reuter developed a program to reduce the sky--noise by eliminating this ``zero--order'' component of the sky--noise, i. e. subtracting the average of simultaneous (here 7) observations from the data. The result of this is shown in the middle panel of Fig. 3, the map is much less noisy but part of the source emission has been subtracted from the map. This effect can be minimized by selecting a model distribution of the source and subtracting the model from the data before calculating the average. As model we took a two dimensional Gaussian fitted in an iterative process to the data. The result is shown on the bottom of Fig. 3. Contour levels are drawn at about 2.5 sigma RMS, there are no negative contours. The reason for including the intermediate step (subtracting the average without correcting by a model) is to show that the result is very little affected by the details of the model. In most cases it will be sufficient to approximate the source by a sum of Gaussians.

  
Figure 3: A map of NGC2146 taking the ``raw'', uncorrected data (top), correcting the data by subtracting the average of simultaneous observations (middle) and by subtracting the average relative to an Gaussian (bottom). The contour levels in all three plots are the same except that the zero level is drawn in the plot shown in the middle of the figure.