Trellis-based coded modulation is the topic of this chapter. Just as in Ungerböck’s and Wei’s papers, based on two-dimensional set-partitioning of M-PSK and M-QAM, the treatment starts from capacity-based gain forecasts. Similar to the Hamming-distance based criterion for good convolutional codes, namely ensuring maximum Hamming distance for incoming and outgoing links to and from a state in the trellis or state diagram, Ungerböck translated this into a criterion based on the squared Euclidean distances. The chapter also follows Wei’s and Ungerböck’s multidimensional set partitioning, especially the famous Wei code that one finds in manifold applications, such as voiceband and DSL modems. Although following the same principle, one-dimensional set partitioning has been delayed to Chap. 22, which discusses Lattices.