Performance modeling of embedded systems is used to explore the system-level design-space and compare a wide range of possible solutions with respect to their performance gain. Calibration and validation of the developed performance models help ensure that the performance predictions of these models are accurate. This paper presents formal definitions of the errors associated with the calibration and prediction process. In addition to a quantitative description of these errors, this paper also provides formal means to estimate their upper bounds, based on known or measured quantities