The power available at the output of solar arrays keeps changing with solar insolation and ambient temperature. Expensive and inefficient, the solar arrays must be operated at maximum power point (MPP) continuously for economic reasons. Of the numerous algorithms for this purpose, perturb and observe (P&O) is a standard. A derivative of gradient ascent method used in the optimization theory, this algorithm introduces a tradeoff between tracking and dynamic performance. This algorithm also has a tendency to drift the system away from the MPP as atmospheric conditions change. With continually changing atmospheric conditions, these inadequacies lead to poor utilization of solar arrays. This paper addresses both the issues. A variable-step-length algorithm is proposed to eliminate the tradeoff. The drift is minimized by evaluating the entire trend in a power versus voltage curve. Analytical results, validated on a prototype system show excellent performance.