Optical long-slit rotation curves have been studied for 304 northern Sb-Sc galaxies from the Courteau-Faber (CF) sample designed for Tully-Fischer (TF) applications; r-band photometry exists also for each galaxy of the sample. The procedure of rotation curve (RC) extraction and construction of optical profiles are analogous to 21cm integrated linewidths. More than 20% of the galaxies were observed twice or more, allowing for a proper determination of systematic errors. Various measures of maximum rotational velocity to be used as input in the TF relation are tested on the basis of their repeatability, minimization of TF scatter, and match with 21cm linewidths. The best measure of TF velocity, V2.2 is given at the location of peak rotational velocity of a pure exponential disk. An alternative measure to V2.2 which makes no assumption about the luminosity profile or shape of the rotation curve is Vhist, the 20% width of the velocity histogram, though the match with 21cm linewidths is not good. The paper shows that optical TF calibrations yield internal scatter comparable to, if not smaller than, the best calibrations based on single-dish 21cm radio linewidths. Tables 6 and 7 contain the values of the parameters fitting the rotation curves of two samples of galaxies: the CF (Courteau-Faber) sample (table6) and the MAT (Mathewson et al., 1992ApJS...81..413M) sample (table7). Two fitting models were used: the ArcTan function (Model 1) where the velocity varies as v(r) = v(0) + K . arctan(R) (R being (r-r_0_)/r_t_) and a multi-parameter function (Model 2) modelling the rotation curve close to a solid-body close to the rotation center, and close to a flat rotation at large distances (see section 5 of the paper).