Three different approaches to minimize control deviations in the case of TMF experiments with high temperature transients (10K/s and higher) will be presented. In the first approach, an empirically derived algorithm is used to calculate a modified temperature command function based on the recorded control deviations. In a second approach, a two-zone model of the plant (specimen, grips, heating and cooling device) is used for control design. The model is first used to design a "soft" PI controller for generating a robust starting output profile. This output profile is then used as a basic reference input profile. The remaining error is minimizes by a "fast" PI controller. A third controller approach is characterized by the determination of an inverse model that is used to define an "ideal" profile of the command variable. This "ideal" profile is then used as a reference input profile.