|
This paper presents the design of two different control approaches for stabilizing the operating point of a microring resonator modulator (MRM) when fast disturbances occur. These arise in particular when the properties of the high-frequency modulation signal change due to the rapid electrical and optical self-heating effects. The challenge lies in the relatively slow dynamics of the heater used as an actuator and the significant limitation of the input. The first approach involves the design of a Model Predictive Control (MPC) suitable for a system like MRM. MPC enables a wide operating range by considering the non-linearities of the system. It can operate in both the stable and unstable regions while accounting for constraints on the control input and states of the system. A drawback for real-time implementation, especially in the case of a fast system like the MRM, is that the computational effort is relatively high in each time step due to the included optimization of the control input over a prediction horizon. On one hand, MPC can function as a benchmark design to showcase achievable control quality in simulation, despite its relatively high computational cost. On the other hand, employing a low-order approximation of the dynamic MRM model allows for offline pre-computation of the MPC internal optimization, thereby significantly reducing the online computational effort. The second approach proposes a computationally efficient PID control with two modes. The first mode is designed for normal operation of the PID controller in a region close to the operating point, while the second mode facilitates returning the system back to the operating point (where the PID controller can be restarted) from outside this region, achieved by applying a feedforward control. The system's state variable, ring temperature, determines the operating regions and the corresponding mode selection. To address the challenge of unknown temperature, an Unscented Kalman Filter (UKF) is designed to estimate the temperature.
|
