Background digital calibration techniques for high-speed, high resolution analog-to-digital data converters

A high-speed, high-resolution analog-to-digital converter (ADC) is a key component in broadband communication transceivers, video imaging systems, and instrumentation. As the ADC speed increases with the advances in IC fabrication technology, the ADC resolution is still limited by the non-ideal effects of the circuits, such as device inaccuracy, component mismatch, and finite device gain. A recent trend for enhancing the resolution is to calibrate the non-ideal effects in background with the aid of digital signal processing. These techniques are preferred since the calibration accuracy is not limited by the accuracy of the analog components, and the calibration tracks the variations of process, voltage and temperature without interrupting ADC's normal operation. This dissertation describes the background calibration techniques for three high-speed, high-resolution ADCs using different architectures: pipelined, floating-point, and continuous-time (CT) [delta]-[sigma]. For pipelined ADCs, a background digital calibration technique with signal-dependent dithering scheme is proposed to overcome the dither magnitude and measurement time constraints with the existing fixed-magnitude dithering. A 15-b, 20-MS/s prototype ADC achieves a spurious-free dynamic range (SFDR) of 98 dB and a peak signal-to-noise plus distortion ratio (SNDR) of 73 dB. The chip is fabricated in 0.18-um complementary metal-oxide-semiconductor (CMOS) process, occupies an active area of 2.3 x 1.7 mm², and consumes 285 mW at 1.8 V. The concept of signal-dependent dithering is also applied to a floating-point ADC (FADC) to calibrate the gain and offset errors in the variable gain amplifier (VGA) stages. A digitally-calibrated 101̃5-b 60-MS/s FADC adjusts its quantization steps instantly depending on the sampled input level and enhances the integral non- linearity (INL) from 24 to 0.9 least significant bit (LSB) at a 15-b level for small input signals. The chip is fabricated in 0.18-um CMOS process, occupies 3.5 x 2.5 mm² , and consumes 300 mW at 1.8 V. In the CT [delta]-[sigma] architecture, the active filter is calibrated by injecting a binary pulse dither and nulling it with an LMS algorithm. The proposed technique calibrates the filter time-constant continuously with crystal accuracy, while the conventional master-slave approaches use additional analog components which limit the calibration accuracy. A 3rd-order 4-b prototype in 65-nm CMOS occupies 0.5 mm² and consumes 50 mW at 1.3 V. It achieves a dynamic range (DR) of 81 dB over an 8-MHz signal bandwidth with a 2.4 Vpp full-scale range. Signal-to-noise ratio (SNR) and SNDR at -1 dBFS are 76 and 70 dB, respectivel