Life time prediction for analog circuit components utilizing digital degradation monitors
Final Report Abstract
The characteristic properties of integrated circuits are influenced by various factors. These include variations due to process variation and changing temperature and voltages during operation, as well as degradation effects due to aging, which lead to a change in circuit properties over time. To ensure circuit functionality, especially in safety-critical applications, monitoring of current circuit performance and degradation may be required. The aim of the LEMON project was to develop a monitoring method for monitoring the degradation of integrated circuits by using the filter coefficients of adaptive filters. The basis of the monitoring developed here is a method for on-line detection and correction of errors within an analog sensor system by digital calibration with adaptive filters. This procedure is extended by the use of a behavioral model of the circuit to be monitored based on response surface models (RSM). This model is used to determine the ideal expected circuit behavior under different operating conditions in order to draw conclusions about the current performance values with the help of the filter coefficients. The filter of the monitoring system consists of two adaptive filters for detecting and correcting offset and gain errors in the output signal of the circuit to be monitored. To determine the filter coefficients, the LMS algorithm is used, which is characterized by its ease of implementation and was identified in the previous work as suitable for use in online correction methods. The adaptive filters require the ideal expected output signal of the circuit as a reference signal for correction. This is provided by the RSM-based behavioral model. The model models the performances to be monitored by the filter as RSMs as a function of the environmental parameters temperature T and supply voltage V based on simulation results and thus represents the ideal behavior. The current operating conditions regarding temperature and supply voltage are measured by sensors. The current performance values of the circuit are determined by weighting the ideal reference perforamnces calculated by the RSMs with the resulting filter coefficients. Initial calibration is also used to determine the individual process corner of the circuit. Monitoring is performed in regular test cycles so that observation and comparison to previous filter coefficients and performance values allow monitoring of degradation. The application of the developed monitoring principle has been evaluated on several example circuits in a simulation-based manner. It is shown that the chosen approach is suitable for determining the current performance under the influence of different operating conditions and process variation. Furthermore, after an initial calibration to capture the initial characteristics, monitoring and detection of degradation due to aging is possible. The general possibility of using the developed monitoring method to determine the current circuit performance was demonstrated using a measurement setup with commercially available operational amplifier circuits.
Publications
-
"Monitoring Analog Circuit Performance using Adaptive Filters and RSM-based Behavioral Models." SMACD/PRIME 2021; International Conference on SMACD and 16th Conference on PRIME. VDE, 2021.
Taddiken, Maike, Steffen Paul & Dagmar Peters-Drolshagen
