The IFS is based on an array of 254x254 hexagonal microlenses --- the so called integral field unit IFU --- located on the adaptive optics focal plane, each microlens having a pitch sampling half the FWHM of the central peak of the diffraction pattern (0.0138 arcsec projected on the sky), so that a total area of 3.5x3.5 arcsec on the sky is sampled. The array of microlenses form an array of micropupils, which are reimaged onto the detector by means of an afocal system made of a dioptric collimator and two dioptric cameras.
An Amici prism inserted in the collimated beam disperses the light from the micropupils at a very low resolution (R~15), with two pixels per resolution element along the dispersion direction. The Amici prism may be replaced by other higher resolution dispersers, combined with narrow band filters. The total number of spectral elements cannot exceed 20. A dichroic beamsplitter feeds two cameras, one for the J-band (0.95-1.35µm), and one for the H-band (1.35-1.75µm). Suitable masks are located before the mirolens array (to occult the central part of the stellar image, to avoid saturation of the detector), on the focal plane (in order to define circular apertures centered on each microlens, reducing the impact of diffraction spikes and of straylight from residual inter-lenses regions) and on the micropupil plane (to reduce straylight). These masks effectively reduce the cross-talk between adjacent spectra at levels of a few 10^-4.