A blind identification algorithm for digital calibration of pipelined ADC

A blind identification based algorithm for digital calibration of pipelined ADC is discussed in this paper. In contrast to traditional approaches that use one highly precise input stimulus to characterize the device, the approach adopted here is based on providing multiple inputs with nonlinearities of separated spectrum to the device-under-test (DUT). A correction code for each output of ADC is determined, which is then used to calibrate the device. Simulation results show that by using the algorithm, trip point error of a 16-bit pipelined ADC with original INL of hundreds of LSBs can be decreased to less than 1 LSB, while input signals of 6-bit linearity are identified to more than 12-bit accuracy. The relaxed requirement of the input signal makes it practical to be generated on-chip and hence is a promising solution for analog and mixed-signal built-in self-test (AMBIST).