6. Conclusions

• With LUMINOUS, it is possible to derive the faint part
  of the luminosity function to the limit of approximately one source per beam.
• The derived solution is not unique, but smoothing the derived LF produces acceptable results.
• When compared with the limit of detection for classical photometry (DAOPHOT was used as a comparison), the derived LF contains about ten times as many stars as the directly detected LF, or when expressed in limiting magnitude, up to two magnitudes fainter for the LF used in this simulation.
• The sampling of the intensities in the image has no effect on the limiting magnitude or uncertainty in the LF, but affects smoothness of the derived LF.
• The uncertainty in the LF grows exponentially with stellar magnitude, faster than the LF, defining an upper limit on the magnitude. The accuracy is dominated by the accuracy of the models of the ADC effect, PSF and read-out noise. The shape of the LF has no significant effect on the accuracy.

Provided that accurate calibrations of the dominating sources of uncertainty are available, LUMINOUS is an excellent complement to classical photometry, allowing the LF to be measured up to two magnitudes fainter, at no additional observational cost.

Acknowledgements

I thank Massimo Stiavelli at Space Telescope Science Institute for many useful discussions on various kinds of noise sources for the WFPC2.