A method of testing figure quality of a convex mirror in which an optical quality substrate material is used. Three separate interferometric null tests are carried out to produce three independent sets of optical path difference (OPD) data. Null lenses, or nulling computer generated holograms (CGHs), are designed and used as needed in each test setup so that spherical aberration is corrected. From the three sets of OPD data, surface figure errors on the rear and front surfaces of the test optic are calculated as well as the OPD error introduced by refractive index inhomogeneity in the substrate material. The rear surface is then corrected, generally using a computer-controlled polishing machine, to reduce rear surface errors to a manageably small level. The front convex surface of the test optic is then corrected to reduce surface figure error to within some specified amount.