Structured illumination microscopy (SIM) is a powerful method for high-resolution 3D-imaging that is compatible with standard fluorescence labeling techniques, as it provides optical sectioning as well as an up to twofold improvement of lateral resolution over widefield microscopy by combining illumination pattern diversity with computational reconstruction. We present a quantitative analysis of the image quality of 3D-SIM using the spectral signal-to-noise ratio (SSNR). In particular, we compare conventional woodpile illumination pattern based 3D-SIM, where the pattern is rotated and translated to acquire the set of raw images that is fed into the reconstruction algorithm, to (square or hexagonal) lattice 3D-SIM, where the pattern is only translated to assemble the input set of raw images. It appears that conventional 3D-SIM has better SSNR than the considered cases of lattice 3D-SIM. In addition, we have also analyzed the impact of the relative amplitude, angle of incidence and polarization of the set of illumination plane waves on image quality, and show how two SSNR derived metrics, SSNR volume and SSNR entropy, can be used to optimize these illumination pattern parameters.