Raman spectroscopy can be used to determine density of gases, because the energy of fundamental vibrational modes is affected by intermolecular distances. The key problem is the estimation of exact peak positions of Raman bands, because the analyses require a precision that is mostly less than the pixel resolution of modern Raman spectrometers. A new method to determine peak positions of Raman bands and atomic emission lines in a discontinuous spectrum without numerical manipulations is tested in this study: modified scanning multichannel technique. Relocation of the gratings with a Sinus Arm Drive can be performed over a distance that is only a fraction of the pixel size that allows peak position estimations with precisions smaller than the pixel resolution and to determine the uncertainty in this estimation. This uncertainty was not determined in previous studies about gas densimeters, resulting in a large variety of inconsistent data. The new method is tested with fluid inclusions in quartz. A CO2 density of 0.1477 ± 0.0006 g·cm−3 and 0.8880 ± 0.0007 g·cm−3 determined with microthermometry correspond to a Fermi dyad of 103.12 ± 0.27 cm−1 and 104.71 ± 0.26 cm−1. A CH4 density of 0.3461 ± 0.0002 g·cm−3 and 0.4011 ± 0.0001 g·cm−3 correspond to peak positions of 2910.66 ± 0.12 cm−1 and 2910.57 ± 0.12 cm−1. The error in these numbers must be regarded as the best estimated uncertainties of peak positions, which are probably slightly adjusted to higher values due to mechanical irregularities of the Sinus Arm Drive in modern Raman systems.