The spectral-density mapping of a 13Calpha-1Halpha vector of Leu10 in the 22-residue peptide hormone motilin [P. Allard, J. Jarvet, A. Ehrenberg, and A. Graslund, J. Biomol. NMR 5, 133-146 (1995)] is extended in this paper to three polarizing fields 9.4, 11.7, and 14.1 T in order to improve the accuracy of the calculated spectral-density function J(omega) and to extend the sampling range up to 750 MHz. The problem with a usually large relative error in J(omegaH) is eliminated since the generally more precise J(omegaH - omegaC) and J(omegaH + omegaC) determined at other fields appear at nearly the same frequencies. The fitting of dynamic models to the points of spectral density was made with error weighting, and the influence of J(omegaH) was found to be negligible. Therefore, the high-frequency part of the spectral-density function is determined essentially without influence from the two transverse-type relaxation rates. In the case of a carbon-proton vector, the relaxation is mainly determined by dipolar interaction and is only weakly influenced by other relaxation mechanisms, which makes it particularly suitable for the spectral-density mapping technique. The measured relaxation rates in the time domain are transformed into the frequency domain by spectral-density mapping, and the slopes in different frequency regions are important parameters when comparing experimental data with theoretical models of motion. Using an adjustable internuclear distance reff, combined with the model-free approach, it is possible to obtain a reasonable fit to measured spectral-density points at J(0) and around J(omegaC). At the same time, however, the high-frequency slope of the spectral-density function defined by J(omegaH - omegaC) and J(omegaH + omegaC) could not be reproduced.