A model for the excitation and emission of SiO masers in evolved stars
, that includes a non-local treatment of the radiative transfer, is pr
esented. The approximations made and their effects, the sources of num
erical noise, and in general the reliability of the model are discusse
d. The code is shown to give an important amount of information on the
theoretical expectations for SiO maser emission. The masers are suppo
sed to be placed in an inner quasi-stationary layer of the circumstell
ar envelope, the properties of which are discussed in view of the avai
lable information. The numerical predictions are generally in good agr
eement with the observational data. In particular, the calculated inte
nsity attains the measured level for a wide range of physical conditio
ns. The maser pump is found to be dominantly radiative, acting through
the DELTAv=1 SiO vibrational transitions. Collisional pump is also po
ssible, but probably gives only a marginal contribution to the total m
aser output. The radiative maser pump naturally explains the observed
dependence of the maser intensity on the stellar infrared continuum, b
oth from object to object and during the stellar variability cycle. Ho
wever some observational features, like the well known low intensity o
f the v=2 J=2-1 line and the possible spatial coincidence of the J=1-0
v=1 and v=2 masers, are not reproduced by the numerical results; poss
ible explanations to this disagreement are discussed. This is the firs
t paper of a series, of which the second part deals with the predicted
spatial extent of the SiO maser emission.