Embedded Model Control allows one to proceed systematically from fine plant dynamics and control requirements to the Embedded Model (EM), which is the core of control design and algorithms. The model defines three interconnected parts: the controllable dynamics, the disturbance class to be rejected and the neglected dynamics. Controllable and disturbance dynamics must be observable from the plant measurements. Control algorithms are designed around the first two parts, while stability and performance are constrained by the third one. The key design issue is discriminating between driving noise and neglected dynamics, to guarantee updating disturbance in view of its rejection. To this end, concept and equations of the 'error loop' are outlined: it maps error sources to performance and shows how to discriminate destabilizing sources, while meeting performance requirements. An introductory example with analytical and simulated results illustrates the design steps.