This paper develops equations for the bias and random noise in the signal estimates from both non-frame transfer CCDs
and frame transfer CCDs having overclock rows that are used to estimate and eliminate the bias due to image smear.
The paper also reports on numerical experiment estimates of signal, bias, and random noise obtained from computer
simulation of the charge generation, charge transfer, and signal processing steps employed with these CCDs to obtain
signal estimates. The theoretical predictions of the exact equations are checked against the experimental results of the
simulation and found to be in close agreement.
Using the analytical formula, the magnitude of the smear bias in the signal estimate for a non frame transfer CCD is
compared to the magnitude of the true minimum signal for CCD operating parameters similar to the ones in a unique
Earth remote sensing application. The bias error due to smear is found to be huge, 1.3 times the magnitude of the true
signal. For the same operating conditions, the total random noises are compared for this CCD and one having overclock
rows to eliminate bias. The random noise of the frame transfer CCD with overclock rows is only 1 electron greater than
the random noise of the non frame transfer CCD. Thus, using the frame transfer CCD with overclock rows to eliminate bias, the additional random noise error incurred is minimal compared to the error from the eliminated bias.