eTULM-MAPCF: Accurate Dynamic Testing of High-Resolution ADCs with Relaxed Source Purity

Coherent sampling and ultra-pure signal sources are two major bottlenecks limiting the mass-production deployment of high-resolution dynamic testing of analog-to-digital converters (ADCs). The principal contribution of this study is the Multi-Amplitude Physically Constrained Fitting (MAPCF) framework, which uses amplitude-dependent constraints to separate signal-source distortion from ADC-induced distortion. MAPCF is coupled with eTULM, an efficient non-coherent-sampling estimator that retains the established FIRE/MIRE two-stage architecture but replaces DFT-peak initialization with closed-form Tabei–Ueda interpolation. Under the evaluated conditions, eTULM reduces the initial frequency error to below 0.1 ppm and requires at most one Levenberg–Marquardt refinement iteration, without zero-padding. Experiments show that eTULM achieves harmonic amplitude estimation errors within 1.2 dB under non-coherent sampling, whereas MAPCF effectively separates signal-source and ADC-induced distortions. For the 16-bit ADS9224R, the separated ADC total harmonic distortion converges to −96.4 dB, consistent with the quantified measurement floor and therefore treated as a lower bound on the intrinsic THD under the evaluated conditions. The framework relaxes the practical signal-source purity requirement under the evaluated testing conditions. The framework enables dynamic testing without coherent frequency tuning under the evaluated conditions, achieves convergence using a single data acquisition, and provides a practical approach for mass-production dynamic testing within the validated testing of ADCs across the envelope.