Quartz crystal resonators are a critical component in many electronic systems, providing the reference frequency source for the system’s clock. However, temperature often affect frequency stability. As a result, frequency-temperature (f-T) characteristic modeling has become an important area of research in frequency control. The traditional f-T modeling method omits system dynamics and can result in significant frequency compensation errors in the case of rapid temperature changes. To address this issue, this paper proposes a dynamic f-T modeling method with considering the thermal hysteresis. A dynamic f-T modeling method based on long short-term memory (LSTM) is presented to reflect the thermal hysteresis characteristics of quartz crystal resonators. Compared to traditional methods, LSTM are suitable for processing and predicting time-series data and consider past temperature history to make predictions. Additionally, transfer learning techniques are used during the training process of the model. Transfer learning fine-tunes the LSTM model for new/unknown crystal readout circuits using less data. Finally, the modeling and testing results on real experimental data show that the proposed method provides better frequency deviation predictions.

This content is only available via PDF.
You do not currently have access to this content.