Research On Wideband Frequency Interleaved Sampling And Reconstruction

With the complexity and diversification of the electronic environment,the frequency and bandwidth of the signal in the environment have been in a high growth state.Coupled with the all-digital development of modern signal processing,the demand for high-speed sampling of broadband signal is unprecedented urgent and vigorous.The sampling rate can be effectively improved by means of multi-channel parallel sampling,such as time-interleaved sampling and reconstruction,frequency-interleaved sampling and reconstruction.However,the system processing bandwidth of the time-interleaved one is equal to the bandwidth of a single channel analog-to-digital converter,leading to the limited acquisition capability of the broadband signals.Whereas the total bandwidth of the frequency-interleaved one is the result of the sum of the bandwidth of all channel analog-to-digital converters,which owns obvious advantages in high-speed sampling scenarios of broadband and even ultra-wideband signals.At the same time,technique of frequency-interleaved sampling and reconstruction also avoids the painful point of the time-interleaved one which is vulnerable to clock mismatch,resulting in a new round of research focus.This paper mainly focuses on the technique of frequency-interleaved sampling and reconstruction and makes theoretical studies on the evaluation of device error,the analysis and calibration of spurious due to channel mismatch error involved in the general and the proposed improved frequency-interleaved sampling and reconstruction model.The simulation experiments provide strong support for the theoretical research results.Meanwhile,the optimal implementation structures of the referred two models are deduced,and the hardware test of the improved frequency-interleaved sampling and reconfiguration model is carried out.Test results verify the effectiveness of the improved model.This paper digs into the key technologies of frequency-interleaved sampling and reconstruction,and strives to provide a driving force for its development and application.1)For the general frequency-interleaved sampling and reconstruction model,two algorithms are proposed to evaluate the actual frequency response of analog sampling filter,including Discrete Fourier Transformation method and cross-correlation calculation method.Both methods can calculate the actual frequency response of the sampling filter according to the only available data output from analog-to-digital converters in the model.The difference lies in that the former method requires a large amount of sample data,while the latter one is able to reduce the amount by introducing additional reference signals and turns out to be much faster.2)The equivalent model of channel mismatch error in the general frequency interleaved sampling and reconstruction model is established,and the expression of system output spectrum with mismatch error is derived.Under the assumption of sinusoidal excitation,qualitative analysis is posed on the spurious effect caused by three kinds of channel mismatch errors,i.e.offset error,gain error and timing error.What's more,according to the theory of poly-phase filtering and the coefficient symmetry of linear phase FIR filter,the low-speed and easy-to-implement structure of reconstruction process in general model is deduced.3)Based on channelization,an improved frequency-interleaved sampling and reconstruction model is proposed.With help of quadrature down conversion and single low-pass filtering,the signal is divided into several uniformly distributed sub-bands and each sub-band is moved to baseband.After quantization,the sampling rate can be recovered by processing of digital up-conversion which also help move the carrier frequency of each baseband signal to its original position.The sole existed digital low-pass filter not only realizes the anti-aliasing function during up-converting,but also ensures the lossless recovery of the output signal.The working mechanism of the improved model is discussed in detail which clarifies that only a unique analog sampling filter and a unique digital reconstruction filter are needed leading to the reduction of the design complexity effectively.4)The equivalent model of channel mismatch error in the improved frequency interleaved sampling and reconstruction model is established,and the independent calculus of three channel mismatch errors is carried out.The analysis results show that the spurious effect of the improved model due to the gain error and the timing error is less than that of the general model.With regard to the calibration of channel mismatch error in the improved model,a calibration method of DC offset error based on Hilbert filter is analyzed.The proposed calibration method does not need to estimate the offset error beforehand and has higher precision.5)An efficient parallel digital up conversion technology is introduced.By decomposing the low-pass filter and the complex mixing signal simultaneously in the traditional digital up-conversion process,the filtering operation and the orthogonal mixing operation can be put before the interpolation operation,thus effectively reducing the data processing pressure before the interpolation.6)The hardware implementation optimization of the improved frequency interleaved sampling and reconstruction model is discussed in depth.On the basis of parallel digital up-conversion technology,an efficient poly-phase DFT implementation model is deduced progressively,which further reduces the operation speed and filter order during the process of reconstruction,and will not impact the characteristic when channel number increases.7)The improved frequency interleaved sampling and reconstruction model is implemented in a broadband IF acquisition and processing module.Meanwhile,the hardware verification is carried out.The hardware architecture of the module and the design of key circuits and algorithms are introduced.The test results show that the improved model is feasible,effective and practical.