An approach to the derivation of variable loop gain sequences of dual-loop digital phase-locked loop (DPLL) [1] is developed based on some modifications of the Kalman filtering formulation, It is shown that optimal loop gain sequences which are independent of measurement noise statistics tan be obtained under a deterministic source model. Computer simulation results demonstrate that the adaptive dual-loop DPLL designed bu? using the proposed method is more robust to noise variations than the adaptive DPLL in [2].