Research On Key Techniques Of Wideband Signal Synthesis Based On BI-DAC

Nowadays,modern electronic systems require that the system signal must be complex wideband signal.However,analog synthesis method is only used to generate simple wideband signals.Digital sampling technique can be used to generate the complex wideband signal,but the complex wideband signal bandwidth is limited by the sampling rate of digital-to-analog converter(DAC).Although the sampling rate of DAC can be directly or indirectly increased by the manufacturing process or using time interleaving(TI)DAC,the output bandwidth is restricted by the zero-order hold(ZOH)characteristic of each sub-DAC in TI-DAC.To break through this bandwidth limitation,bandwidth interleaving digital-to-analog converter(BI-DAC)is thus proposed.There are some the system internal errors(time delay error,phase offset and aliasing errors)in BI-DAC,and the complex wideband signal generated by BI-DAC is distorted due to these errors.To improve the spectrum quality of the complex wideband signal generated by BI-DAC when achieving high bandwidth,this thesis studied the principle of the complex wideband signal synthesis in BI-DAC.Further,the generation principle of time delay error,phase offset and aliasing errors in BI-DAC were also studied.Finally,this thesis focused on these errors' compensation,and the main research work were described as follow:(1)Principle analysis of wideband signal synthesis and errors generation:In the frequency domain,we studied the principle of wideband signal synthesis of an ideal M-channel BI-DAC.Then,we studied the generation principle of time delay error,phase offset and aliasing error which deteriorated the spectrum of the complex wideband signal generated by a non-ideal BI-DAC.(2)Estimation and compensation methods of time delay error and phase offset:We studied the auto-power spectrum expression of the complex wideband signal generated by a non-ideal dual-channel BI-DAC in the overlapped frequency band,and a linear phase function including time delay error and the phase offset was derived."Three-point" method was used to unwrap this linear phase function to simultaneously estimate time delay error and phase offset.According to the estimation values,digital predistortion technique was used to predistort the input signal of sub-DAC such that time delay error and phase offset were simultaneously compensated.The experimental results showed that the absolute error estimation accuracy of delay error and phase offset reached 5.77%and 0.75%,respectively.And the simultaneous compensation of delay error and phase offset is effectively realized.(3)Minimax design of digital finite impulse response(FIR)filters using linear programming(LP):We deduced the transfer function and approximate error function of a non-ideal M-channel BI-DAC.Based on LP,minimax method was used to minimize the approximate error so that we obtained the optimal coefficients of digital FIR filters which were used to decompose the frequency band.The aliasing errors caused by the non-ideality of analog mixers and filters were thus effectively compensated.Taking a four-channel BI-DAC as experimental example,using our proposed design,the maximum aliasing error was reduced to-61 dB and the computational complexity was 2.5728×105.These two key parameters are superior to the traditional WLS design and the optimal design based on the bi-conjugate gradient stabilized method.(4)Low computational complexity minimax design of digital FIR filters using LP:On the basis of minimax design of digital FIR filters using LP,a new order optimal method was proposed to optimize all digital FIR filters' orders such that low computational complexity minimax design of digital FIR filters using LP was presented.Comparing with traditional minimax design of digital FIR filters using LP,this low computational complexity design could not only achieve the same aliasing error,but also reduce the computational complexity.Taking a four-channel BI-DAC as example,using our proposed low computational complexity design,the maximum aliasing error was reduced to-62.1 dB.Further,the computational complexity was 1.8304×105 which was lower than that of minimax design using LP.(5)Minimax and weighted least squares(WLS)designs of digital FIR filters using second-order cone programming(SOCP):The transfer function and approximate error function of a non-ideal M-channel BI-DAC were deduced.Based on SOCP,minimax and WLS methods were used to minimize the approximate error for obtaining the optimal coefficients of digital FIR filters which were used to decompose the frequency band.Unconstrained minimax and WLS designs of digital FIR filters using SOCP were thus proposed to compensate the aliasing errors generated by the non-ideality of analog mixers and filters.The linear equality constraints and the convex quadratic inequality constraints were studied,and they were added to unconstrained minimax and WLS designs using SOCP for proposing constrained minim.ax and WLS designs using SOCP.Take a four-channel BI-DAC as experimental example.The maximum aliasing errors of unconstrained minimax and WLS designs of digital FIR filters using SOCP were reduced to-73.9 dB and-80.5 dB,respectively.Their computational complexity were 9.66×105 and 8.31744×108,respectively.In contrast,unconstrained WLS design of digital FIR filters using SOCP achieved better aliasing errors reduction but larger computational complexity.In the selected frequency band(?),the maximum aliasing errors of constrained minimax and WLS designs of digital FIR filters using SOCP were lower than-110 dB.Out of this frequency band,the maximum aliasing errors achieved by these two constrained designs were-36.5 dB and-42.9 dB,respectively.Additionally,the smaller the width of the selected frequency band,the smaller the maximum aliasing error within the frequency band,and the larger the maximum aliasing error out of the frequency band.Moreover,the computational complexity of constrained WLS design of digital FIR filters using SOCP was larger than that of constrained minimax design of digital FIR filters using SOCP.