Available bandwidth knowledge is very useful to network protocols. Unfortunately, available bandwidth is also very difficult to measure in packet networks, where methods to guarantee and keep track of the bandwidth (eg, weighted fair queuing scheduling) do not work well, for example the Internet. In this paper we are dealing with an available bandwidth sampling technique based on the observation of packet time dispersion in a packet train or pair. In standard techniques the available bandwidth is sampled by using a "bytes divided by dispersion" (or "bytes over time", BoT) calculation and then filtered. This method of calculating available bandwidth samples has been used in all packet dispersion related work. We propose a new sampling method of available bandwidth called ab-probe. The ab-probe method uses an intuitive model that helps understand and correct the error introduced by the BoT sample calculation. We theoretically compare the new model with the previous one, exploring their differences, observability and robustness. We argue that the model may significantly improve protocols that can use an available bandwidth measurement, in particular transport-level protocols that currently use the BoT calculation.