CLEAN

Often, the sensitivity of the observation run will not be limited by system noise, but by the consequences of incomplete sampling in the Fourier transform plane. Obvious contributors may be sidelobes (from the type of grading function used for the Fourier components), grating rings (from the Fourier response to a regular spacing of the antenna tracks), or aliasing (e.g. from interpolation of data onto a regular, rectangular grid, to allow use of a Fast Fourier Transform).

One procedure to overcome defects introduced by undersampling the () plane is the CLEAN algorithm (Högbom, 1974), which has the following steps:

1. Start with the raw data "dirty" map (in , plane).
2. Go to the highest peak on the map, subtract from it the "dirty" beam (the uncorrected response of the array to a point source). Remember the intensity and the position.
3. Go to the next highest peak on the new map, and subtract the dirty beam from this peak too.
4. Iterate, as desired.
5. When the noise level is reached, add back gausssians at the positions derived in the previous steps.

The result is a "cleaned" map, consistent with the observed data. (But not necessarily the true distribution on the sky).