Design and analysis of multi-gigahertz track and hold amplifiers

Ultra high speed, moderate resolution data acquisition systems such as high-end test and measurement equipment, radio, radar require ultra high speed analog to digital converters(ADCs). The track and hold amplifier (THA) is a crucial front-end building block in ADC since it makes a great impact on the dynamic performance of an ADC. At ultra high sampling rates the achievable resolution of a THA decreases as the input frequency increases. Particularly, as presented in the literature, the resolution is far less than 8 bit when the sampling frequency reaches 15 GSample/sec (GS/s), which cannot meet our design goal of an 8 bit, 15 GS/s THA. Many techniques can be used to increase the resolution. Among them two effective approaches are to reduce noise and distortion in the THA. In order to suppress the noise and distortion, first, a noise and distortion analysis of an ultra high speed THA operating in the saturation region is investigated. Simplified transistor models that are justified and valid for CMOS and SiGe HBT are derived for both noise and distortion analysis respectively. Theoretical noise analyses for both track mode and hold mode of the THA show that, at ultra high frequency, the noise contribution from the parasitic capacitors is significant. Simulation results locate the major noise contributor and indicate that the dominant noise in the THA is KT/C thermal noise. Volterra series analysis is applied to analyse the weakly nonlinear behavior of the THA. It shows strong agreement between the Volterra analysis and SpectreRF simulation. Several techniques are then applied to decrease the effects of noise and distortion in the THA. These techniques involves open-loop linearization used to compensate for the nonlinear distortion, the replica switch technique employed to reduce the hold mode feed through as well as the parasitic capacitance compensation technique applied to expand the input bandwidth. As a result, the linearity of the proposed THA is significantly