A Data-Driven Approach for Indirect Estimation of INLmax_abs Using Dynamic Parameters as Input Features

Abstract INL testing based on the code-density histogram requires a large number of samples, which significantly degrades testing efficiency. To address this issue, this paper proposes a data-driven indirect estimation approach for INL max_abs . In particular, it takes dynamic parameters such as SNR, SINAD, THD, SFDR, and harmonics as input features, and constructs a machine learning model to estimate INL max_abs . For model construction, a Hybrid KNN-SVR model is developed by integrating the local learning capability of KNN with the global fitting advantage of SVR, thereby improving robustness across datasets with different distributions. The experimental results show that the RMSE values for INL max_abs estimation on the five ADC datasets are 0.0532, 0.0607, 0.0301, 0.2006, and 0.2083 LSB, respectively. Furthermore, the Hybrid KNN-SVR model demonstrates performance comparable to or even better than complex models such as XGBoost and ANN. This work provides a valuable basis for the rapid assessment of INL performance.