Date of Award
Doctor of Philosophy
Electrical and Computer Engineering
Analog-to-digital converters (ADC) are one of the most important components in modern electronic systems. In the mission-critical applications such as automotive, the reliability of the ADC is critical as the ADC impacts the system level performance. Due to the aging effect and environmental changes, the performance of the ADC may degrade and even fail to meet the accuracy requirement over time. Built-in self-test (BIST) and self-calibration are becoming the ultimate solution to achieve lifetime reliability. This dissertation introduces two ADC testing algorithms and two ADC built-in self-test circuit implementations to test the ADC integral nonlinearity (INL) and differential nonlinearity (DNL) on-chip.
In the first testing algorithm, the ultrafast stimulus error removal and segmented model identification of linearity errors (USER-SMILE) is developed for ADC built-in self-test, which eliminates the need for precision stimulus and reduces the overall test time. In this algorithm, the ADC is tested twice with a nonlinear ramp, instead of using a linear ramp signal. Therefore, the stimulus can be easily generated on-chip in a low-cost way. For the two ramps, there is a constant voltage shift in between. As the input stimulus linearity is completely relaxed, there is no requirement on the waveform of the input stimulus as long as it covers the ADC input range. In the meantime, the high-resolution ADC linearity is modeled with segmented parameters, which reduces the number of samples required for achieving high-precision test, thus saving the test time. As a result, the USER-SMILE algorithm is able to use less than 1 sample/code nonlinear stimulus to test high resolution ADCs with less than 0.5 least significant bit (LSB) INL estimation error, achieving more than 10-time test time reduction. This algorithm is validated with both board-level implementation and on-chip silicon implementation.
The second testing algorithm is proposed to test the INL/DNL for multi-bit-per-stages pipelined ADCs with reduced test time and better test coverage. Due to the redundancy characteristics of multi-bit-per-stages pipelined ADC, the conventional histogram test cannot estimate and calibrate the static linearity accurately. The proposed method models the pipelined ADC nonlinearity as segmented parameters with inter-stage gain errors using the raw codes instead of the final output codes. During the test phase, a pure sine wave is sent to the ADC as the input and the model parameters are estimated from the output data with the system identification method. The modeled errors are then removed from the digital output codes during the calibration phase. A high-speed 12-bit pipelined ADC is tested and calibrated with the proposed method. With only 4000 samples, the 12-bit ADC is accurately tested and calibrated to achieve less than 1 LSB INL. The ADC effective number of bits (ENOB) is improved from 9.7 bits to 10.84 bits and the spurious-free dynamic range (SFDR) is improved by more than 20dB after calibration.
In the first circuit implementation, a low-cost on-chip built-in self-test solution is developed using an R2R digital-to-analog converter (DAC) structure as the signal generator and the voltage shift generator for ADC linearity test. The proposed DAC is a subradix-2 R2R DAC with a constant voltage shift generation capability. The subradix-2 architecture avoids positive voltage gaps caused by mismatches, which relaxes the DAC matching requirements and reduces the design area. The R2R DAC based BIST circuit is fabricated in TSMC 40nm technology with a small area of 0.02mm^2. Measurement results show that the BIST circuit is capable of testing a 15-bit ADC INL accurately with less than 0.5 LSB INL estimation error.
In the second circuit implementation, a complete SAR ADC built-in self-test solution using the USER-SMILE is developed and implemented in a 28nm automotive microcontroller. A low-cost 12-bit resistive DAC with less than 12-bit linearity is used as the signal generator to test and calibrate a SAR ADC with a target linearity of 12 bits. The voltage shift generation is created inside the ADC with capacitor switching. The entire algorithm processing unit for USER-SMILE algorithm is also implemented on chip. The final testing results are saved in the memory for further digital calibration. Both the total harmonic distortion (THD) and the SFDR are improved by 20dB after calibration, achieving -84.5dB and 86.5dB respectively. More than 700 parts are tested to verify the robustness of the BIST solution.
Chen, Tao, "Built-in self-test and self-calibration for analog and mixed signal circuits" (2019). Graduate Theses and Dissertations. 17161.