The small satellite ‘Galileo Galilei’ (GG) will test the universality of free fall and hence the weak equivalence principle which is the founding pillar of general relativity to 1 part in 1017. It will use proof masses whose atoms differ substantially from one another in their mass energy content, so as to maximize the chance of violation. GG will improve by four orders of magnitude the current best ‘Eöt-Wash’ tests based on slowly rotating torsion balances, which have been able to reach their thermal noise level. In GG, the expected violation signal is a relative displacement between the proof masses of ≃ 0.6 pm caused by a differential acceleration aGG ≃ 8 × 10−17 ms−2 pointing to the center of mass of the Earth as the satellite orbits around it at νGG ≃ 1.7 × 10−4 Hz. GG will fly an innovative acceleration sensor based on rapidly rotating macroscopic test masses weakly coupled in 2D which up-converts the signal to νspin ≃ 1 Hz, a value well above the frequency of natural oscillations of the masses relative to each other νd = 1/Td ≃ 1/(540 s). The sensor is unique in that it ensures high rotation frequency, low thermal noise and no attenuation of the signal strength (Pegna et al 2011 Phys. Rev. Lett. 107 200801). A readout based on a very low noise laser interferometry gauge developed at Jet Propulsion Laboratory (≃ 1 pm Hz−1/2 at 1 Hz demonstrated) allows the short integration time to be fully exploited. A full scale sensor with the same degrees of freedom and the same dynamical features as the one to fly in GG has been setup on ground (GGG). The proof masses of GGG are affected by acceleration and tilt noise acting on the rotating shaft because of ball bearings and terrain microseismicity (both absent in space). Overall, by means of appropriate 2D flexure joints, these noise sources have been reduced by a factor almost 105 down to a differential acceleration between the proof masses of ≃ 7 × 10−11 m s−2 (at 1.7 × 10−4 Hz up-converted by rotation to ≃ 0.2 Hz). The corresponding noise in the relative displacements of the proof masses, read by co-rotating capacitance bridges, is ≃ 180 pm, which is 300 times larger than the target in space. GGG error budget shows that it can reach a differential acceleration sensitivity aGGGgoal ≃ 8 × 10−16 m s−2, not limited by thermal noise. This value is only a factor 10 larger than what GG must reach in space to meet its target, and slightly smaller than the acceleration noise of the torsion balance. It can be achieved partly by means of weaker joints and an optimized mechanical design—so as to improve the attenuation factor—and partly by replacing the current ball bearings with much less noisy air bearings (also used in torsion balance tests) so as to reduce input noise. A laser gauge readout with noise level rlaser-ro ≃ 30 pm Hz−1/2 at 0.2÷3 Hz will be implemented.