The inversion of bimodal dynamic light scattering data is very difficult. Tikhonov regularization method is the often used inversion algorithm, however, the influence of different regularization matrices on the inversion is not clear yet. Two bimodal particle size distributions with 6 levels of noise are inverted by using identity matrix L1, first order difference matrix L2 and second order differential matrix L3. Simulation data shows that the bimodal resolution decrease with the increase of noise level. The anti-interference ability of the algorithm is stronger when the components of bimodal distribution are closer. Under the same noise level, the bimodal resolution of matrix L3 is of the best, while the error of inversion is minimum; correspondingly the bimodal resolution of matrix L1 is of the worst, while the error of inversion is maximum. The matrix L3 can distinguish the smallest peak value size ratio, while matrix L1 can only distinguish the biggest peak value size ratio. Under the same noise level, peak value size ratio is bigger and the bimodal resolution is stronger. Therefore, matrix L3 should be used in order to get the correct inversion result by inverting the noisy scattering data. Finally, the inversion of experimental particles confirms this conclusion.