Exoplanet direct imaging instruments such as the VLT-SPHERE and the Gemini Planet Imager will soon be in operation, providing a quantum leap in comparative exoplanetary science. A similar leap will again happen when the next generation of such instruments is built for the future extremely large telescopes gaining a factor of 4-5 in spatial resolution and up to several hundred in sensitivity. These instruments are built around extreme adaptive optics (XAO) systems providing correction both of atmospheric turbulence and inevitable aberrations and wave front ripples due to the optical surfaces. Their performance is however limited by the non-common path aberrations (NCPA) due to the differential wave front errors existing between the XAO sensing path and the science path, leading to residual speckles that hide the faint exoplanets in the coronagraphic image. Accurate calibration of the NCPA is mandatory in order to correct these quasi-static speckles and allow observation of exoplanets, 10 to 10 times fainter than their host star. Several approaches to NCPA calibration are currently being developed, including different incarnations of the Zernike phase contrast method. We here present work on one such sensor, known as ZELDA (Zernike sensor for Extremely accurate measurements of Low-level Differential Aberrations). Building on previous experience with phase mask-based wave front sensors for segment phasing and with phase mask coronagraphs, ZELDA proves to be highly promising both as an upgrade path for current-generation instruments and for implementation into future-generation systems, both ground-based and space based. We present results obtained on a high-contrast test bed where the ZELDA sensor performs on-line calibration of an AO loop feeding a coronagraph system. The setup is representative of systems such as SPHERE, hence providing clear evidence for the applicability of the concept as an upgrade path for that instrument. We a Kjetil.dohlen@lam.fr Third AO4ELT Conference Adaptive Optics for Extremely Large Telescopes Florence, Italy. May 2013 ISBN: 978-88-908876-0-4 DOI: 10.12839/AO4ELT3.13376 Adaptive Optics for Extremely Large Telescopes III also present an error budget indicating that an improvement in image contrast of a factor of at least an order of magnitude with respect to the currently expected performance is within reach. The implementation of such an upgrade would give a second youth to the instrument, by significantly increasing its discovery space.