Tiny compensation of pressure drift measurements due to long exposures to high temperatures

Many industrial, medical, automotive, and consumer applications increasingly demand highly accurate pressure sensors to operate reliably. However, a variety of real-world conditions, between which prolonged exposure to high temperatures, cause sensor accuracy to drift. This article proposes a very tiny neural network (NN)-based compensation system that correct the drift measurement in real time. The system is capable of compensating for drift accuracy from 1.68 to 0.1471 hPa, reducing the variance from 0.0115 to 3.5x10-4 in worst case conditions. To evaluate the integration of the system into the sensor digital logic, a hardware accelerator has been designed for the processing element required to run the proposed NN. By using CMOS standard cells with 130 nm TSMC technology, synthesis reports an area occupation of 0.0373 mm2 and a power consumption of 1.07 μW. The integration of the system into sensor circuitry will allow for a more robust and reliable intelligent pressure sensor