Understanding the effects of Single Event Transient Errors (SETs) and Single Event Upsets (SEUs) in Convolutional Neural Networks (CNNs) is crucial for their deployment in space applications. However, this task is challenging due to the lack of explainability in neural networks and the probabilistic nature of error injection. This work presents a high-performance frame-work for analyzing event errors in quantized neural networks using dedicated TensorFlow delegates. Furthermore, we conduct a detailed analysis of an illustrative small-size network. The results demonstrate significant differences in accuracy degradation depending on the affected layer and suggest that the effect of the affected bit can be modeled with a dual binary type model.