Raman spectroscopy is a very popular, non-destructive tool for the structural characterisation of carbons. Raman scattering from carbons is always a resonant process, in which those configurations whose band gaps match the excitation energy are preferentially excited. Any mixture of sp 3 , sp 2 and sp 1 carbon atoms always has a gap between 0 and 5.5 eV, and this energy range matches that of IR-vis-UV Raman spectrometers. The Raman spectra of carbons do not follow the vibration density of states, but consist of three basic features, the G and D peaks at approximately 1600 and 1350 cm - 1 and an extra T peak, for UV excitation, at ~980-1060 cm - 1 . We propose to rationalise the vast range of experimental data available in literature at any excitation wavelength by a simple model, which considers the main factors influencing the Raman spectra. The great advantages of multi-wavelength Raman spectroscopy will be clarified by a series of examples. In particular we show how it can be used to probe the structural changes induced by annealing and by nitrogen introduction. UV Raman spectroscopy also probes heteropolar σ bonds in a complementary way to infrared spectroscopy. We demonstrate the direct detection of C H vibrations in hydrogenated DLC samples, Si H and Si C vibrations in amorphous silicon and amorphous silicon-carbon alloys and the easier probe of CN sp bonds in amorphous carbon nitrides.