Research On Frequency-Interleaved Sampling Via Zero-IF Architecture

High-speed analog-to-digital converters(ADCs)are in high demand for accurate sampling of emerging wideband radio frequency signals ranging from 5G wireless net works,medical imaging,Terahertz radar,and aerospace.Current state-of-the-art high speed serial links,mm-wave imagers,and oscilloscopes necessitate ADCs with sample rates greater than 50GS/s and resolutions exceeding 6 bits.To reach a broader band width,those sophisticated ADCs usually leverage parallelism,which is what both the time-interleaved(TI)and frequency-interleaved(FI)architectures do.However,the accu racy of TI-ADCs may be compromised by sampling clock jitters and reduced bandwidth due to enlarged overall input impedance with an increased number of parallel-connected ADCs,which entail sample-and-hold(S/H)circuits of higher bandwidth and precision Therefore,most TI-ADCs sampling above 20GS/s have resolutions less than 5 bitsFI-ADC offers the opportunity to relax the bandwidth requirement on samplers and reduce the clock jitter susceptibility.This is achieved by channelizing the wideband spec trum into multiple equally spaced subbands,which are then frequency-translated to base band,before sampled-and-held concurrently at a lower rate.It features large bandwidth high speed,medium-to-high resolution with respectable jitter robustness,albeit suffer ing from circuit impairments such as spectral leakages,LO harmonic interferences,in phase/quadrature(I/Q)and aliasing images.To solve these issues,the main contributions of this dissertation are fivefold1)A frequency domain system behavioral model of an M-channel FI-ADC is pre sented.The model addresses mismatch errors due to imperfect channel separation,in complete anti-aliasing,I/Q branch imbalances,and jitter-induced distortions.Two sim plified mathematical models,namely direct-sampling and baseband-sampling are estab lished for TI-and FI-ADC systems respectively and their dynamic performances are dis cussed.Based on this,their ENOB responses and jitter robustness are compared respec tively2)Krylov subspace algorithms are introduced for improving the synthesis filter de sign in conventional FI-ADCs.A equalization model in matrix form is formulated which transforms the mismatch compensation problem into a linear system.Several Krylov methods are compared and a Gauss-Seidel preconditioner is utilized for further optimiz ing the proposed design approaches.The autocorrelation matrix of magnitude response is introduced to quantize the spectrum leakage and evaluate the performance of overlapping cancelation between the nearby subbands3)The bandwidth and jitter robustness benefits of FI over TI come at a cost of area due to the use of I/Q modulation,which generally demands intensive lowpass filtering(LPF)for decent channel separation and anti-aliasing,making a fully integration compli cated.A simpler FI-ADC prototype obliviating such dedicated LPFs(i.e.,LPF-less)is proposed to relax this tradeoff.In this novel framework,channel recombination is per formed before calibration.It circumvents the needs for analog harmonic rejection and lowpass filtering,thus enabling significant simplification of the complex analog frontend design4)A comprehensive system behavioral model of LPF-less FI-ADCs is derived The model addresses all the aforementioned linear circuit imperfections simultaneously.Given error parameters or their distributions,closed-form expressions of the expected output signal-to-noise ratio and image rejection ratio are formulated with an arbitrarynumber of channels.The LO and sampling clock jitters and combined overlapping ef-fects including overlapping between channels,aliasing within and across channels are also discussed.For the first time,this work addresses the harmonic-folding issue in mul tichannel FI-ADCs digitally,which requires only one analog mixer per signal branch5)A discrete-time equivalent model and its matrnx form for M-channel LPF-less FI-ADC systems are derived.The cascaded 2M-periodically time-vary(PTV)system model is also formulated.Therefrom,an integrated compensation scheme based on re laxation iteration is proposed that contrasts existing signal recovery approaches which calibrate specific types of errors separately.Three iterative correction structures,namely Richardson,Jacobi and Gauss-Seidel,as well as their convergence conditions are studied Two implementation methods of PTV compensation filters are proposed and their hard ware complexities are discussed.The performance of this novel framework is compared against conventional solutions using extensive simulations.Very favorable numerical re sults are observed.