The most important noise source for both the WFPC and WFPC 2 is the readout noise. This is a random noise with an approximately Gaussian distribution which occurs while reading out the CCD and measuring the charge deposited in each pixel. For WFPC the noise is about 13 electrons RMS per pixel, which is well-sampled by the 7.5 electron per DN digital conversion. For WFPC 2 the readout noise is only about 5 electrons per pixel, and is undersampled by both the 7.5 and 15 electrons per DN digital conversions. This undersampling of the noise may impact deconvolution and statistical analyses of WFPC 2 data.
WFPC suffers from an electronic problem in its ``sample and hold'' circuitry which corrupts the analog-to-digital conversion of the data. Basically, as various bits in the A-to-D conversion are tested, the bits sometimes fail to set properly so that the resulting digital values are too low. This adds a small noise which is typically about 1/2 DN, but can be as large as 2 DN, depending on the pixel value. This again may impact statistical analyses on the data. WFPC 2 has a similar problem, but at about twenty times lower level.
For most targets, WFPC images must be ``preflashed'' with faint lamps to eliminate charge transfer errors during the readout. Images are typically preflashed to about 30 electrons, which introduces about 1 DN of noise per pixel. At present it is believed that WFPC 2 images will not require any preflash.
Table 2 summarizes the effective total noise and other factors related to faint
target detection for both WFPC and WFPC 2 cameras. Fig. 5 compares observed
WFPC, Richardson-Lucy deconvolved WFPC, and simulated WFPC 2 images of a field
containing faint targets. While the PSF wings are reduced in the deconvolved
image, noise prevents detection of faint objects easily seen by WFPC 2.
