We study the possibility to undo the quantum mechanical evolution in a time reversal experiment. The naive expectation, as reflected in the common terminology ("Loschmidt echo"), is that maximum compensation results if the reversed dynamics extends to the same time as the forward evolution. We challenge this belief, and demonstrate that the time tr for maximum return probability is in general shorter. We find that tr depends on λ = ε evol /εprep, being the ratio of the error in setting the parameters (fields) for the time reversed evolution to the perturbation which is involved in the preparation process. Our results should be observable in spin-echo experiments where the dynamical irreversibility of quantum phases is measured.In this Letter we study the probability of return P (t 1 , t 2 ) for a generalized wavepacket dynamics scenario. The system is prepared in some initial state Ψ prep , which can be regarded as the outcome of a preparation procedure which is governed by a Hamiltonian H prep . We assume that the quantum mechanical evolution is generated by Hamiltonians with classically chaotic limit: The state is propagated for a time t 1 using a Hamiltonian H 1 , and then the evolution is time-reversed for a time t 2 using a perturbed Hamiltonian H 2 . The corresponding evolution operators are U 1 and U 2 . The probability of return to the initial state isThere are two special cases that have been extensively studied in the literature. The traditional wavepacket dynamics scenario [1] is obtained if we set t 2 = 0. In this context the "survival probability" is defined asThe "Loschmidt echo" (LE) scenario is obtained if we set t 1 = t 2 = t. In this context the "fidelity" is defined asThe theory of the fidelity was the subject of intensive studies during the last 3 years [2,3,4,5,6,7,8,9,10]. It has been adopted as a standard measure for quantum reversibility following [2] and its study was further motivated by the realization that it is related to the analysis of dephasing in mesoscopic systems [11].In the present Letter we consider the full scenario of a time reversal experiment. The probability to find the system in its original state is P (t) = P (t, 0) before the time reversal (t < T /2), andafter the time reversal (t > T /2). The period T is the total time of the experiment. The naive expectation, which is also reflected in the term "Loschmidt echo", is to have a maximum for P (t) at the time t = T . We are going to show that this expectation is wrong. We find that the maximum return probability is obtained at a time t r which in general is shorter than that. Namely,If we have t r = T /2 we say that there is no reversibility. If we have t r ∼ T we say that we have a nearly perfect echo. We show that t r /T is a function of a dimensionless parameter 0 < λ < ∞. Namely,where ε prep quantifies the difference between the evolution Hamiltonian H and the preparation Hamiltonian H prep , while ε evol quantifies the difference between the two instances H 1 and H 2 of the evolution Hamiltonian, which are used for ...