Compressive spectral video (CSV) has emerged as a novel research topic in computer vision, having powerful implications in retrieving a four-dimensional tensor from a reduced set of two-dimensional compressed measurements. The advances in CSV architectures have shown that the recovery of four-dimensional data is possible, but the quality is still far from the requirements of real applications. To overcome these limitations, this work analyses the parameter design of a novel dynamic spectro-temporal encoding methodology based on a digital micromirror device sync with a tunable filter.