A real-time algorithm for estimating the subsample delay between two noisy sinusoidal signals is presented. The new technique is based on the recently proposed discrete-time quadrature delay estimator (QDE). The algorithm uses all of the in-phase and quadrature-phase components of the received signals to produce a bias-free estimate of the delay between the input signals. The technique directly updates the delay error estimate, which is in turn used to adapt the coefficients of a simple fractional delay filter (FDF). The algorithm is insensitive to variations in the amplitudes of the input signals, and does not require an accurate estimate of the frequency of the input sinusoids. The technique is simple to implement and has a significantly reduced complexity compared to other adaptive techniques. Simulations show that in the presence of system noise, the new estimator outperforms conventional fractional delay filter based estimators.