In order to detect directly faint planets around other stars it is important to develop techniques to cancel the central starlight. We report progress in terms of null depth on the visible nulling experiments at JPL, essential for NASAs future space missions as well as for an upcoming sounding rocket based experiment. A level of 1 Million to 1 nulling ratio (1e-6 rejection ratio) has been obtained for the first time for visible laser light, and a level of 100000 to 1 nulling ratio has been obtained for a relatively large (5%) bandpass at 650nm. Using a combination of a fiber array and deformable mirror a null of 1e-6 would correspond to a contrast of 1e-9 at a distance of 2-3$\lambda$ in the airy disk plane, only 10x away from TPFs goal. The configuration used was a fiber fed Mach Zender type interferometer, using 2 or 3 mirrors in each two arm, placed in an enclosed air container. For broadband light we used tilted dispersion glass plates in the nuller arms and an avalanche photon diode module to cover the large dynamic range. We describe a variety of conditions that have to be met and optimized to reach very deep nulls and sub-nanometer optical path difference stability, such as optical alignment, symmetry in the two arms, mechanical stability and vibration isolation.