The Background Of The Pipelined Adc Digital Correction Algorithm

The speed and resolutions of ADC almost have gone to utmost in the CMOS process if only rely on analog circuits design. Using calibration to compensate and correction errors is an important technique to improve the performance of ADCs. Calibration is a significant way to break the limitation of CMOS process in the current integrated circuits design technology.In the traditional designs of Pipelined ADCs, seveval errors will degrade the performance of the ADC. Among these errors, the memory effects have aroused researcher's attention, it may instead the gain-error and capacitor-mismatch and become the dominant error under special condition.Two new calibration algorithms are proposed that correct for memory errors by digital post-processing of the ADC output. One is based on the FIR algortithm while another is based on the LMS algorithm.First, the sources of memory effects are analyzed, and the memory parameter is mesured under the condition of SMIC 0.18μm and the structure of opamp-sharing. The influence of memory effect on the 12-bit 100MS/s Pipelined A/D is also researched. Then, all the signals are made the Z-transform. After a series of equation analysis, a direct conlusion is obtained that the second memory effects is the dominant error. It is in accordance with the simulation results.The FIR algorithm is relatively simple. According to the memory parameter, the digital outputs are multiplied by a special number in the way of simple displacement.The LMS algorithm could not only correct the memory effects but also the gain error of the first stage. A calibration signal is injected into stage input signal and processed simultaneously, which is controlled by pseudo random code. For sake of calibration signal, the memory effects parameter can be measured, through which the calibration coefficient is adjusted actively, and then the gain errors will be corrected at digital domain background.For the FIR method, the steady-state SFDR and SNR has been improved 37 dB and 16 dB respectively. The INL are less than 1.5 LSB. For the SAR method, the steady-state SFDR and SNR has been improved 39 dB and 21 dB respectively. The INL are less than 0.5 LSB. The MATLAB simulation results show that system linearity is improved by the calibration.