Artificial Neural Network Based Digital Temperature Compensation Method For Aluminum Nitride MEMS Resonators

This paper reports on the demonstration of an effective use of an artificial neural network (ANN) algorithm for the implementation of a digital temperature compensation method (DTCM) for aluminum nitride (AlN) MEMS resonators. This method resulted in an improved frequency-temperature stability (14 ppm from -40 degrees C to + 80 degrees C with respect to 100 ppm when using a resistive-feedback control circuit), while consuming very low ovenization power (as low as 390 mu W over the same temperature range). To our knowledge, this implementation exhibits the highest figure of merit (product of oven gain times temperature range and 1/power consumption) ever demonstrated for any ovenized MEMS resonators. Interestingly, the same technique can be used to compensate for fabrication-induced frequency variations, hence eliminating the need for trimming.