A large list of common proper motion stars selected from the third Astronomischen Gesellschaft Katalog (AGK3) was monitored with the CORAVEL (for COrrelation RAdial VELocities) spectrovelocimeter, in order to prepare a sample of physical binaries with very wide separations. In paper I,66 stars received special attention, since their radial velocities (RV) seemed to be variable. These stars were monitored over several years in order to derive the elements of their spectroscopic orbits. In addition, 10 of them received accurate RV measurements from the SOPHIE spectrograph of the T193 telescope at the Observatory of Haute-Provence. For deriving the orbital elements of double-lined spectroscopic binaries (SB2s), a new method was applied, which assumed that the RV of blended measurements are linear combinations of the RV of the components. 13 SB2 orbits were thus calculated. The orbital elements were eventually obtained for 52 spectroscopic binaries (SBs), two of them making a triple system. 40 SBs received their first orbit and the orbital elements were improved for 10 others. In addition, 11 SBs were discovered with very long periods for which the orbital parameters were not found. It appeared that HD 153252 has a close companion, which is a candidate brown dwarf with a minimum mass of 50 Jupiter masses. In paper II, 80 wide binaries (WBs) were detected, and 39 optical pairs were identified. Adding CPM stars with separations close enough to be almost certain they are physical, a "bias-controlled" sample of 116 wide binaries was obtained, and used to derive the distribution of separations from 100 to 30,000 au. The distribution obtained doesn't match the log-constant distribution, but is in agreement with the log-normal distribution. The spectroscopic binaries detected among the WB components were used to derive statistical informations about the multiple systems. The close binaries in WBs seem to be similar to those detected in other field stars. As for the WBs, they seem to obey the log-normal distribution of periods. The number of quadruple systems is in agreement with the "no correlation" hypothesis; this indicates that an environment conducive to the formation of WBs doesn't favor the formation of subsystems with periods shorter than 10 years.