Research On Signal Generation Based On Band Decomposition And Error Calibration

The rapid development of electronic information industry exhibits a high demand for wideband signal generation.Arbitrary waveform generator(AWG)which is an indispensable component of modern electronic test and measurement system,becomes more and more difficult to meet the requirements on wideband test.This is mainly due to the limited sampling rate of single digital-to-analog convertor(DAC)chip.The current solution based on time interleaved DAC(TIDAC)has limited ability to improve the output bandwidth.The sampling and hold characteristics of DAC still restrict the bandwidth with in two Nyquist band.Another promising way to enlarge the bandwidth is the structure based on band decomposition.This dissertation mainly focuses on the band decomposition based on hybrid filter band(HFB).Comparing with another two band decomposition schemes(the discrete Fourier transform based and the digital down converter based),the HFB structure has advantage in resource cost and is suitable for real-time generation.However,the HFB system suffers from the aliasing error if the analysis filters and synthesis filters are not well matched.Besides,the nonideal characteristic of circuits will also significantly degrade the system performance.Hence,this dissertation mainly focuses on the design and calibration of band decomposition based on HFB,the main works are as follows:(1)On the basis of analyzing the current signal reconstruction method with single DAC,a signal generation method based on the analog band decomposition is introduced.The influence caused by the DAC sampling and hold characteristic is discussed.Then,a signal generation method based on the combination of HFB and mixers is proposed.The principle is explained,and the perfect reconstruction condition is formulated by the derivation of system transfer function.(2)A method of designing digital analysis filter bank for HFB is proposed.The perfect reconstruction problem is transformed into a set of linear equations and then a filter coefficients calculation method based on biconjugate gradient stabilized method is introduced.Bandwidth weighting factor and distortion error weighting factor is introduced in this method to realize the artificial control of reconstruction performance.In this situation,the transform matrix may become ill-conditioned.To solve this problem,a Gauss-Seidel pre-conditioner method is introduced and hence lead to a faster converge speed.(3)On the basis of analyzing the design method of synthesis filter bank and comparing the reconstruction performance when using different type of synthesis filter banks,an optimization method of synthesis filters is proposed.With this method,the distortion error and aliasing error will be further attenuated without increasing the analysis filter length.(4)A filter bank design method based on the second-order cone optimization is introduced.In addition,an optimization model with parameter constrains is proposed.With this method,one can adjust the distortion performance and aliasing performance separately which helps to simplify the design produce.(5)The HFB system suffers from the nonideal characteristics of subchannels.In this dissertation,the offset error,gain error and timing error is analyzed respectively.The system transfer function with error is derived.Taking sinusoid input as example,the locations of spurious components are summarized.An offset error estimation method based on the statistic characteristic is proposed.For the gain error,an estimation method based on the principal component analysis is introduced and an iteration method is proposed to estimate the timing error.(6)The equivalent discrete-time model and the periodic time-varying system model is derived.Based on this,a digital pre-distortion based on the relaxion method is proposed.This pre-distortion method could calibrate the channel errors comprehensively.Taking Richardson iteration and Gauss-Seidel iteration as example,the principle of error calibration and the produce of iteration is explained.Implementation of the pre-distortion filter is proposed.