We report new metallicities for stars of Galactic globular cluster M 4 using the largest number of stars ever observed at high spectral resolution in any cluster .
We analyzed 7250 spectra for 2771 cluster stars gathered with the VLT FLAMES+GIRAFFE spectrograph at VLT .
These medium resolution spectra cover by a small wavelength range , and often have very low signal-to-noise ratios .
We attacked this dataset by reconsidering the whole method of abundance analysis of large stellar samples from beginning to end .
We developed a new algorithm that automatically determines the atmospheric parameters of a star .
Nearly all data preparation steps for spectroscopic analyses are processed on the syntheses , not the observed spectra .
For 322 Red Giant Branch stars with V \leq 14.7 we obtain a nearly constant metallicity , < \textrm { \mbox { $ [ { Fe } / { H } ] $ } } > = -1.07 ( \sigma = 0.02 ) .
No difference in the metallicity at the level of 0.01 \textrm { dex } is observed between the two RGB sequences identified by ( 58 ) .
For 1869 Subgiant and Main Sequence Stars V > 14.7 we obtain < \textrm { \mbox { $ [ { Fe } / { H } ] $ } } > = -1.16 ( \sigma = 0.09 ) after fixing the microturbulent velocity .
These values are consistent with previous studies that have performed detailed analyses of brighter RGB stars at higher spectroscopic resolution and wavelength coverage .
It is not clear if the small mean metallicity difference between brighter and fainter M 4 members is real or is the result of the low signal-to-noise characteristics of the fainter stars .
The strength of our approach is shown by recovering a metallicity close to a single value for more than two thousand stars , using a dataset that is non-optimal for atmospheric analyses .
This technique is particularly suitable for noisy data taken in difficult observing conditions .