Various radiation effects can be anticipated, based on similarities to other CMOS devices. These expectations, coupled with the performance characteristics described in the previous section drive the choices of radiation testing.

In simplified terms, charge trapping involves the trapping of radiation induced holes in the oxide. This occurs in both field oxides and gate oxides, and results in a negative shift in the threshold voltage. The amount of charge trapped, and hence the threshold shift, is a function of electric field, temperature, transistor polarity and radiation type. Annealing after radiation may release the trapped holes, undoing some of the threshold shift.

The simplest effect of trapped charge is just a threshold voltage shift, tending to turn n-channel transistors on and
p-channel transistors off. The effects on the circuit can be predicted easily by simulation, with reference to Figure 2. An important consequence of this threshold shift is a change in the source follower load currents. This can, of course, be avoided by use of current mirrors to supply the VLN and VLP voltages. The DC output levels and dynamic range are also affected, but the changes are not simple to estimate, due to the opposing effects on the source followers and loads. Also, if the threshold shift varies from transistor to transistor, the offset FPN will be increased.

As the threshold voltage of n-channel transistors decreases with dose, their subthreshold leakage increases. A more serious problem, for non-hardened processes such as this, occurs when the threshold shift results in inversion of the field oxide or bird’s beak region around n-channel transistors. The resulting source-drain leakage will increase the supply current and eventually cause functional failure. At lower levels, it may affect the functioning of the column multiplexer. Even the lowest level leakage would cause problems in the pixel, probably manifesting itself in noise and nonlinearity. It was for this reason that the design measures in Section II-D were taken.

The generation of interface states is an even more complicated a subject than hole trapping. Due to their energy states in the silicon bandgap, they function as generation-recombination centers, with states near mid-gap being the most effective. When interface states are located in a depletion region, they act as sources of dark current. Increased dark current is expected to be the dominant radiation effect in active pixel sensors, leading to increased shot noise. Variability in the added dark current will also lead to increased FPN.

The depletion region in a photodiode device, such as the CPS32, intersects the surface only around the periphery of the diode, so the dark current due to interface states is expected to scale with the perimeter of the device. In contrast, in a CCD or photogate APS the depletion region covers the entire detector area. Thus, photodiode devices may be expected to have a lower total dose induced dark current than photogate devices. The advantage of this choice diminishes for small pixels, however, as the perimeter-to-area ratio increases with decreasing pixel size.

Like hole trapping, the formation of interface states depends on electric field, temperature and transistor type, with interface state densities reportedly lower for p-channel device than for n-channel devices [7]. This was the primary reason for the choice of a p-channel pixel. Interface states are also reported to concentrate in regions of high mechanical stress, such as at the edge of field oxide, leading to the choice of the fully enclosed diode design.

Unlike trapped holes, interface states can not generally be annealed out at device-compatible temperatures. Indeed, during annealing, the process of removing trapped holes appears to result in the creation of interface states. This causes a "reverse annealing" effect in CCDs, where the dark current increases with annealing after radiation [9]. A similar effect is expected for active pixel sensors.

Protons, in addition to producing total dose damage in the oxide, also create displacement damage in the bulk silicon. The rate of displacement production is a fundamental property of the interaction between high energy protons and silicon, and is therefore not very sensitive to technological variations. As a result, displacement damage effects in CCDs can be extrapolated to APS.

Displacement damage in the depletion region produces dark current, and a damage factor of 2.8.10^-11 nA/p⁺ has been reported for CCDs irradiated with 10 MeV protons [10]. This can be usefully expressed as 0.05 nA/cm²/krad, since 10 MeV protons are a good representative of a shielded proton spectrum. A similar value may be expected for the APS. Loss of red response due to displacement damage in the neutral region is not a problem for the CPS32, sing the collection volume is small.