Effective Gain Analysis and Statistic Based Calibration for Ring Amplifier With Robustness to PVT Variation

A 2.5-GS/s Four-Way-Interleaved Ringamp-Based Pipelined-SAR ADC with Digital Background Calibration in 28-nm CMOS

“…More importantly, ringamps are more suitable for fine CMOS processes and have application scenarios in various types of ADCs due to the versatility of their structures. In the last decade, many ringamp-based ADC works [24,[29][30][31][32][33][34][35][36][37][38][39] have appeared. Essentially, there are two different directions of the application, one for high resolution (signal-to-noise-and-distortion-ratio (SNDR) ≥ 70 dB) and the other for high speed (sampling rate fs ≥ 500 MHz), such as the dual-deadzone RAMP-based two-step SAR ADC [34], which achieves the highest SNDR for a ringampbased high-resolution ADC, while [33] used a dead zone degeneration technique to realize the fastest sampling rate (fs = 1 GHz) for single-channel implementation.…”

“…In this work, we have chosen the LVT transistor as the third stage, to achieve high bandwidth. We have improved the bias-enhanced ringamp with PVT compensation and fast-start circuits based on our previous work in [35]. Figure 2b shows the conceptual bode plot of this self-biased and bias-enhanced ringamp.…”

“…The half-amplification structure used in the 2nd-stage further alleviates the requirement for the 2nd-stage amplifier. More details about the channel ADC can be obtained from [35]. The output data are captured by the on-chip RAM (FIFO, which will be explained in Section 3.3) without decimating the evaluation of the "real" performance of the channel ADC.…”

“…The output data are captured by the on-chip RAM (FIFO, which will be explained in Section 3.3) without decimating the evaluation of the "real" performance of the channel ADC. A statistics-based calibration method is adopted, as in [35], to calibrate the nonlinear gain error of the ringamp for the channel ADC off-chip. After this, the mismatches of the TI-ADC are calibrated in the background, as discussed in Section 3.4 in the off-chip using Matlab.…”