Accuracy of atmospheric parameters of FGK dwarfs determined by spectrum fitting

We performed extensive tests of the accuracy of atmospheric parameter determination for FGK stars based on the spectrum fitting procedure Spectroscopy Made Easy (SME). Our stellar sample consists of 13 objects, including the Sun, in the temperature range 5000--6600~K and metallicity range -1.4 -- +0.4. The analysed stars have the advantage of having parameters derived by interferometry. For each star we use spectra obtained with different spectrographs and different signal-to-noise ratios (S/N). For the fitting we adopted three different sets of constraints and test how the derived parameters depend upon the spectral regions (masks) used in SME. We developed and implemented in SME a new method for estimating uncertainties in the resulting parameters based on fitting residuals, partial derivatives, and data uncertainties. For stars in the 5700--6600 K range the best agreement with the effective temperatures derived by interferometry is achieved when spectrum fitting includes the H$\alpha$ and H$\beta$ lines, while for cooler stars the choice of the mask does not affect the results. The derived atmospheric parameters do not strongly depend on spectral resolution and S/N of the observations, while the uncertainties in temperature and surface gravity increase with increasing effective temperature, with minima at 50~K in Teff and 0.1~dex in log g, for spectra with S/N=150--200. A NLTE analysis of the TiI/TiII and FeI/FeII ionisation equilibria and abundances determined from the atomic CI (NLTE) and molecular CH species supports the parameters we derived with SME by fitting the observed spectra including the hydrogen lines.