A study on the on-line system identification and PID tuning of a buck converter

In this paper we study the on-line system identification process and the proportional-integral-derivative (PID) tuning of a buck converter. The system identification process was performed using a recursive least squares algorithm. The estimation error and parameter error were generated to demonstrate that the system was converging to its true parameters. The estimation error shows an absolute value of approximately 1 × 10-5 in less that 10ms. All the parameters were effectively converging in less that 100μs. Once the system was properly identified, an offline PID controller was designed to further implement it on the adaptive loop. Three different techniques were used to satisfy the requirements of the buck converter: phase and gain margin, pole-zero cancellation and frequency loop shaping. Phase and gain margin still prevails as the easiest method to design controllers. Pole-zero cancellation is based on pole-placement and is fairly easy to implement in order to obtain the gains of a PID controller. However, although these controllers can be easily designed, they do not provide the best response compared to the Frequency Loop Shaping (FLS) technique in terms of frequency and time responses.