Summary form only given: The discontinuous Galerkin Time Domain (DGTD) methods, such as DG-SETD and DG-FETD, are shown to be effective in the simulation of multiscale, transient problems. It can solve large problems by dividing the computational domain into several subdomains, thus the large system matrix is transformed into a few moderate-sized matrices. The computation resource required for solving these matrices is much less than solving the original large one. The numerical fluxes are employed to fulfill the energy communication between adjacent subdomains.In the traditional DGTD methods for Maxwell's Equations, the variables E and H are used, as the curl-conforming basis functions are naturally suitable for the numerical fluxes which use the tangential components of the fields on the interface. A new scheme of DG-FETD method is developed, which use the field variables E and B. Compared to the traditional method that requires different orders of interpolation for E and H, this new method can suppress the spurious modes with the same order interpolation polynomials for both E and B. Thus, this new method is shown to be advantageous in both memory consumption and CPU time. In this work, the DG-SETD and the DG-FETD based on the variables E and B are combined to solve multiscale problems. This new hybrid DG-SETD/FETD method maintains the capability of mesh flexibility from the previous hybrid SETD/FETD method when dealing with the multiscale problems. The advantages of less DoFs to suppress the spurious modes from the introduction of EB scheme guarantee the newly proposed method faster and less expensive in computation load. Numerical results show that it is more advantageous over the traditional EH scheme in almost all aspects, and thus can replace the latter for multiscale problems.