Can we trust SET Injection Models?

As the small structures on contemporary ASICs become increasingly sensitive to radiation particle hits, it is important to understand the related effects. Experimental studies referring to this require high efforts, therefore analog-level simulations have become the method of choice. While it is comparatively easy to represent the actual circuit in a simulator like SPICE at any desired level of detail – including parasitic effects, e.g. – the representation of the particle hit is the crucial problem. The state-of-the-art approach is to apply a current pulse with double-exponential shape to a transistor by attaching a current source across its channel terminals. In another, more refined but less popular model two current sources are used, each connected from bulk to one of the channel terminals. In this paper we apply these two different models for the simulation of a Muller C-element and try identify, how much difference the choice of the model for the particle hit actually makes. Our results show that even for this relatively simple target there are already remarkable mismatches in the model predictions, which clearly confirms the need for further research into SET simulation models.