Abstract: Bender element (BE), resonant column (RC) and cyclic torsional shear (TS) tests are performed to measure the small strain shear modulus of dry sands. The characteristics of the received signals in BE tests are investigated, and different interpretation methods are used to determine the travel time of shear waves. The results indicate that it is important to account for the compliance of the testing system in RC tests, especially when the specimen is stiff. In BE tests, the near field component arrives first and its initial polarization is opposite to that of the input signal. The near flied decreases as the input frequency increases, but it always exists. In practice, it is recommended that the input frequency should increase continuously until the output signal becomes stable. Note that the amplitude of the first arrival of the shear waves may be much smaller than that of the succeeding ones. The travel time can be reliably determined by the start-start method in the time domain, however, it significantly depends on the input frequency by the peak-peak, the cross correlation and the cross power methods. The comparison indicates that the BE, RC and TS tests can obtain consistent results on clean sand, and the consistency does not depend on the particle size.