The fluorescence yield is a key ingredient in cosmic ray energy determination. It is sensitive to pressure, temperature and humidity. Up to now the fluorescence yield of the brightest line at 337 nm has been measured in an absolute way in one set of conditions, whereas fluorescence yields at the other wavelengths have been relatively measured for different conditions. Thus, absolute calibration for all the lines is unclear. We will do all measurements at once using the same apparatus: all the lines will be measured absolutely and not relatively for all conditions. For that we will use the 3-5 MeV electron beam of the PHIL accelerator (Photon Injector at LAL), shooting in a box filled with air at varying pressures, temperatures and humidity. Delta rays resulting from the beam collisions with Nitrogen are responsible for the light yield. The light detection probability should be independent of its emission point especially at the delta ray stopping point. The idea is to use an integrating sphere, encapsulated in a vessel where pressure, temperature and humidity can be varied. This sphere will have two ports for the beam (in and out), one more port dedicated to a NIST photodiode for calibration and another port feeding optical fibers going to: A) a grating spectrometer equipped with cooled CCD. B) a photomultiplier with BG3 filters to measure directly the integrated yield. Calibrations at the percent level, will give each line spectrum yields with a precision between 2 to 5%. A special issue will be to estimate the leakage due to "high energy" delta rays. Thus, we the air density will be increased, the beam energy will be lowered until the beam stops inside the sphere. Then, the energy loss will be precisely derived from the Bethe-Bloch formula. We will present the set-up.