The overall efficiency (including atmospheric transmission, reflectivity of the telescope mirrors and overall efficiency of the optics and of the detector) for broad-band imaging should be about 25% in H and K, and about 15% in J. The main noise source for both broad- and narrow-band imaging should be the photon noise associated to the background+source flux. The K-band background will be dominated by the thermal emission from the telescope primary and secondary mirrors; therefore, the K-band background will depend on the dome temperature and on the K-band reflectivity of the mirrors coating (that is not know yet, but should be around 85%); a conservative estimate of the K-band background is 12 mag arcsec-2 . Instead, in H and J band the background emission is dominated by the OH atmosferic lines; the intensity of these lines depends on the atmospheric conditions: background as high as 12.5 mag/arcsec2 in H and 14.5 mag/arcsec2 in J are conservative. Table 4 lists the expected limiting magnitudes for broad and narrow band imaging for a 10 sigma detection, in 1 hour of integration on the source. Flat fielding (and possibly background subtraction) might increase the noise by a factor 2½, and will therefore lower the limiting magnitude by ~ 0.4. Also, in several cases the frame for flat fielding and sky subtraction must be obtained off-source, that will require an integration time longer by a factor of 2.
The sensitivity of the low-resolution spectroscopic mode will probably be limited by the statistical photon noise associated to the combined flux from background and source, although the exact limiting magnitude will depend on the intensity of the OH lines in the specific spectral region. In Table 4 we report the expected average limiting magnitudes for the various low-resolution spectroscopic modes.
The expected performances of the high-resolution mode are difficult to quantify, as the overall efficiency of the Si-echelle grism is not yet measured. At a resolution of ~ 104 the dominant noise in the J and H band will probably be the pixel read-out noise, as the OH sky lines can be filtered out.