Extrasolar planet observation and characterization is a major objective for astronomy in the coming decade. The
Spectro-Polarimetric High contrast imager for Exoplanets Research (SPHERE) instrument for the Very Large
Telescope (VLT) is currently under development to achieve this objective with greater performances than any
existing instruments. It will have the capability to observe planets down to few masses of Jupiter at separations as
small as 0.1 arcseconds from the central star thanks to dedicated extreme adaptive optics system and coronagraph.
Planets characterization will be achieved with IRDIS, one of the 3 science modules of SPHERE, using its long slit
spectroscopy (LSS) mode along with a Lyot coronagraph at low (~40) or medium (~400) resolution. Similarly to
the dual-band imaging (DBI) technique dedicated to detection, the main limitation of the long slit spectroscopy
is the speckles generated by instrumental static and quasi-static aberrations, especially when observing very close
to the star. Another performance limitation is the chromatic dependence of the speckles in the field, known as
speckle chromatism, which creates a modulation of the spectrum. Finally, we are also facing problems related
to long slit spectroscopy, such as slit efficiency, thermal background and difficulty to preserve the spectrum
continuum. Intensive simulations have been performed using the CAOS framework under IDL, in order to
simulate realistic spectra in terms of signal, noise and instrumental artifacts. We present here in details the
expected performances among with a detailed analysis of the various noise sources contributions. This analysis
leads to a quantitative estimation of the characterization limitations and optimized data reduction procedures.