A method and apparatus for producing thin lamella for TEM observation. The steps of the method are robust and can be used to produce lamella in an automated process. In some embodiments, a protective coating have a sputtering rate matched to the sputtering rate of the work piece is deposited before forming the lamella. In some embodiments, the bottom of the lamella slopes away from the feature of interest, which keeps the lamella stable and reduces movement during thinning. In some embodiments, a fiducial is used to position the beam for the final thinning, instead of using an edge of the lamella. In some embodiments, the tabs are completed after high ion energy final thinning to keep the lamella more stable. In some embodiments, a defocused low ion energy and pattern refresh delay is used for the final cut to reduce deformation of the lamella.