In general, the accuracy is the same as the tables in Chapter 6 of the WFPC2 Handbook, or about 5 to 10%.
For narrow band filters and stellar sources, results can be less accurate (typically around 20% accuracy) if narrow stellar absorption features interact with the filter bandpasses.
For spectral lines, both narrow lines (monochromatic assumption) and broad filters should have good accuracy (5 to 10%). If broad lines are observed with narrow filters, such that the filter throughput varies across the line, then the accuracy is reduced. We suggest using SYNPHOT in STSDAS to get better accuracy in situations with complex spectra / filter interactions.
IMPORTANT NOTE: The algorithm used for point source calculations assumes optimal weighting of pixels in the PSF during point source photometry (e.g. PSF fitting). In cases where the SNR is limited by background noise (i.e. readnoise, dark current noise, or noise from sky emission), simple aperture photometry can give much lower SNR values than predicted by ETC, especially if large apertures are used (i.e. with areas much larger than 10 WF pixels, or 20 PC pixels). For large apertures in the background noise limited case, aperture photometry will give an SNR of approximately:
SNR(aperture) = SNR(ETC optimal) * 0.17 / R for WFC, or SNR(aperture) = SNR(ETC optimal) * 0.11 / R for the PC, where SNR(ETC optimal) is the ETC result, and R is the aperture radius in arcseconds.
Some effects are currently ignored. Cosmic-ray splitting is not included in the "Final Result," though there is detailed discussion of cosmic-ray splitting farther down the output page. For exposures where the noise is dominated by read out, the exposure splitting will significantly increase the noise (e.g. increase by root(2) for 2-way splitting). Information in the "intermediate results" output section can be used to judge the importance of the read out noise. Also, chip-to-chip differences in the read noise are ignored (6% effect). For the sky backgound, only a mean zodiacal contribution is used; sky background variations (up to ~40%) can occur if bright Earth light scatters into the telescope.