Font Size: a A A

Pipelined Adc Behavioral Modeling And Digital Calibration Algorithm Research

Posted on:2011-07-16Degree:MasterType:Thesis
Country:ChinaCandidate:Y ZhengFull Text:PDF
GTID:2208360308966192Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Recently, as a hot topic in mixed-signal circuit field, high-speed and high-accuracy pipeline ADC is widely applied in various occasions, such as wireless communication and mobile multimedia. With the downscaling of CMOS process and reduction of voltage supply, it is more difficult to design high-gain OPAMPs than before. Given other analog errors, digital calibration algorithm is regarded as a key to improve accuracy.First of all, behavioral model of pipeline ADC is designed on Simulink to verify the performance of digital calibration algorithms on different error conditions. Afterwards, based on the general calibration principle, the scheme that uses reference codes from slow-but-accurate ADC to calibrate fast-but-inaccurate ADC is discussed.To improve accuracy of slow ADC, the offline algorithm is transplanted from 1 bit/stage to 1.5 bits/stage for eliminating comparator's offset errors better. Redundancy stages make it run backgroundly to track the change of errors, with gain compensation algorithm to enhance code linearity in the process of renewing the weigths.Next, calibration structures and weight estimation based on least mean square (LMS) algorithm are derived to eliminate various typical errors. Non-linear terms, proved feasible theoretically to mitigate OPAMPs'non-linearity effect, are added and the weights are estimated, using the LMS algorithm from back-end stages to front ones recursively. In order to renew weights backgroundly, two paralleled LMS algorithms that minimize the distance from reference codes and maximize code linearity are designed respectively. Some other important problems such as down-sampling multiples, synchronization of slow and fast clocks and elimination of front-end SHA circuit are also analyzed.For 14-bit ADC, SFDR is enhanced from 56 dB to 99 dB after calibrated by the offline algorithm while INL and DNL decrease from 50 LSB and 1.3 LSB to 1.4 LSB and 0.4 LSB respectively. When downsampling multiple is 10 and non-linear effect of OPAMPs generates maximum compression of 5% at full-scale output, SFDR of 12-bit fast ADC with only linear-term calibration is improved from 40 dB to 75 dB using reference codes from offline-calibrated ADC. Non-linear term makes SFDR up to 85 dB.
Keywords/Search Tags:pipeline ADC, behavioral model, digital calibration algorithm, least mean square, non-linear effect of OPAMP
PDF Full Text Request
Related items