2results about How to "High stability requirements" patented technology

The utility model discloses a 25G clock data recovery instrument which is composed of a clock extraction circuit and a high-speed signal input interface. The clock extraction circuit can accurately extract a clock signal from a 25G high-speed data signal, and a phase comparator, a voltage-controlled oscillator and a loop filter are utilized to track the phase change of the signal and generate a stable local clock synchronous with the data rate. The problem that a clock is easily interfered in high-speed data transmission is solved, a stable reference is provided for data processing and transmission, and the bit error rate is reduced. The high-speed signal input interface receives 25G data streams and works cooperatively with the clock extraction circuit. The stable local clock ensures that all links are accurately synchronized when data are received and processed, the data are processed according to a correct time sequence, the data processing efficiency is improved, the data integrity is ensured, and the system can meet application scenes such as 5G communication and high-speed network transmission which have high requirements for speed and stability.

The present application relates to a resonance sensor reading self-adaptive optimization method and system based on reinforcement learning, comprising the steps of: integrating the current sampling spectrum local maximum value set, the historical sampled spectrum local maximum value point set and its linear compensation trend to construct a state space vector, and taking it as the input of the reinforcement learning strategy network; fusing the action based on the strategy network output, the fusion action including the historical local selected maximum value point selection action and the corresponding compensation action, and the fusion action being used to simultaneously realize the spectrum peak value determination and the frequency correction; the present application can fully utilize the historical information and dynamic characteristics of the sensor by introducing the modeling method based on reinforcement learning, realize the real-time optimization of the reading, and overcome the problem that the traditional fixed filtering or single point correction method is not sensitive to interference; the fusion action mechanism adopted simultaneously considers the peak value selection and the frequency compensation, and significantly improves the anti-noise and anti-vibration ability of the resonance sensor in complex environment.