A diffuser has an interesting texture, as you may expect. I think the attempt is to make a "normal distribution" of cell sizes and random spatial distribution, to make as uniformly random as possible the diffusing action.
These are all shown first for BD Plan in brightfield, then darkfield, and finally with polarizers. When available, M Plan brightfield, polarized and DIC is then presented.
The first two images of the final row are two 100x images with M Plan objectives, and show the effect of restricting the field stop to limit stray light. The second two are different polarizer angles with DIC.