Digital Cancellation of Analog Mismatch Noise in Pipelined ADCs

本研究的目的是通过开发一种数字信号处理技术来纠正模拟电路不匹配，从而扩展流水线模数转换器（adc）目前的性能限制。随着软件可重构性和数字信号处理的发展趋势，流水线adc已广泛应用于蜂窝电话基站和宽带有线调制解调器等应用中。然而，模拟元件不匹配限制了目前制造的流水线adc的精度，使其低于许多通信应用中所需的精度。例如，在目前转换速率在10MHz或以上的流水线adc中，模拟元件不匹配所施加的精度上限约为75dB峰值信号噪声失真（SINAD）。近年来，人们提出了几种不同的失配校准技术来解决失配问题，但到目前为止，它们还没有突破上述75dB SINAD上限。问题是，迄今为止提出的校准技术最终受到其模拟元件精度的限制，随着转换率的增加，其精度往往会降低。本研究涉及开发一种新的流水线ADC失配校准技术，以避免此问题。新技术与其他失配校准方案的不同之处在于，它在ADC正常工作期间连续测量和消除ADC输出中由元件失配引起的噪声；在A/D转换之前，不需要特殊的校准信号或自动校准阶段。失配噪声的测量和消除都完全使用数字逻辑进行，但与之前提出的数字校准技术不同，该技术不需要额外的管道级或每级比特。该研究的目标是：1)完善该技术；2)通过仿真和理论分析量化其性能限制；3)开发CMOS流水线ADC原型，通过该技术实现创纪录的性能，作为原理验证。

The purpose of this research is to extend the present performance limits of pipelined analog-to-digital converters (ADCs) through the development of a digital signal processing technique that corrects for analog circuit mismatches. With recent trends toward software reconfigurability and digital signal processing, pipelined ADCs have become widely used in applications such as cellular telephone base-stations and wideband wireline modems. However, analog component mismatches have limited the accuracy achieved by present fabricated pipelined ADCs to levels below those desirable in many communications applications. For example, the accuracy ceiling imposed by analog component mismatches in present pipelined ADCs with conversion rates at or above 10MHz is approximately 75dB peak signal-to-noise-and-distortion (SINAD). Several distinct mismatch calibration techniques have been proposed in recent years to address the mismatch problem, but so far they have not made it possible to break through the above-mentioned 75dB SINAD ceiling. The problem is that the calibration techniques proposed to date are ultimately limited by the precision of their analog components, which tends to decrease as the conversion rate is increased. This research involves the development of a new pipelined ADC mismatch calibration technique that avoids this problem. The new technique differs from other mismatch calibration schemes in that it continuously measures and cancels noise in the ADC output arising from component mismatches during normal operation of the ADC; no special calibration signals or auto-calibration phase are required prior to A/D conversion. Both the measurement and cancellation of mismatch noise are performed entirely using digital logic, but unlike previously proposed digital calibration techniques the technique does not require additional pipeline stages or bits per stage. The objectives of the research are to 1) refine the technique, 2) quantify its performance limits through simulation and theoretical analyses, and 3) develop a CMOS pipelined ADC prototype that achieves record-setting performance enabled by the technique as a proof-of-principle.