We present the results of a cosmic microwave background (CMB) anisotropy ground-based millimetric experiment sensitive to fluctuations on angular scales of ~2 deg. Four independent bands centered at 3.3, 2.1, 1.3, and 1.1 mm collected ~550 hr of observation during the summer of 1994. The instrument was located on the island of Tenerife at an altitude of 2400 m. The low water vapor content and the atmospheric stability of the site, combined with new techniques to subtract atmospheric noise, result in the reduction of atmospheric contamination in the lowest frequency channel to a level of ~1.5 times the instrument noise. Detailed estimations of Galactic foreground contamination show that this contribution is negligible at |b| >~ 12 deg. Two different multipole bands (l=53^{+22}_{-15} and 33^{+24}_{-13}) are analyzed showing that our technique to subtract the atmospheric contribution is more effective in the multipole band at l=53 . A likelihood analysis of these data reveals the presence of a common signal between the channels at 3.3, 2.1, and 1.3 mm, corresponding to band-power estimates of [l bar (l bar +1)Cl/(2 pi )]^{1/2}=2.0^{+1.0}_{-0.8}x10^{-5} and [l bar (l bar +1)Cl/(2 pi )]^{1/2}=4.1^{+2.4}_{-2.2}x10^{-5} for the l=53 and 33 multipole bands, respectively. Calibration uncertainty has been treated as a systematic effect. The level of fluctuations in the l=53 band is in good agreement with our preliminary analysis, with measurements by other experiments working at similar angular scales, and with the predictions of standard cold dark matter (CDM) models.