Research On Key Technologies Of Broadband Compressed Sensing Receiver

To solve the contradiction between the lack of spectrum resources and the low utilization of spectrum,cognitive radio technology puts forward the requirements of wideband spectrum sensing and reconfigurable multi-band reception for the receiver,which makes ADC face the severe challenge of ultra-high sampling rate.The wideband compressed sensing receivers compressed sample the ultra wideband spectrum and reconstruct the signal with compressed sensing theory,which remove the redundant information of the sampling sequence while alleviating sampling pressure of the ADCs.However,there are some problems in the hardware-implemented compressed sensing receivers,such as the deviation of the sensing matrix from the calculated value,the low reconstruction rate of the classical compressed sensing algorithms and the high cost of hardware implementation.To correct the non-ideal characteristics of the hardware,this dissertation proposes a sensing matrix measurement method based on the synchronous wideband modulation signal.Then,for the high column correlation sensing matrix,a Double Screening Orthogonal Matching Pursuit(DSOMP)algorithm is proposed to improve the reconstruction rate of the support set.Finally,a new compressed sampling structure,Alternate Modulation Wideband Converter(A-MWC),is proposed to further reduce the requirment of the total sampling rate for blind reconstruction.The main contents and innovations of this dissertation are summarized as follows:1.Based on Modulated Wideband Converter(MWC),a compressed sensing receiver with a dynamic range of 60 dB is designed.The receiver can sense the frequency range of DC-3 GHz.Without prior information of the carriers,it can simultaneously receive 6 wideband real signals arbitrarily distributed in the working band,and the bandwidth of each signal can reach up to 40 MHz.The receiver has 6 sampling branches whose ADC works at 240 Msps,which achieves compressed sampling of up to 6 Gsps Nyquist sampling rate.The receiver is at the leading level in terms of sensing frequency band range,receiving frequency band number,signal bandwidth,and dynamic range,and so on.2.To correct the non-ideal characteristics of the hardware,a sensing matrix measurement method based on the synchronous wideband modulation signal is proposed.Due to the non-ideal characteristics of the hardware,the weight of spectrum folding in the compressed sensing receiver deviates from the calculated sensing matrix seriously,which can not lead a satisfactory signal reconstruction.Because of the high cost of calibrating devices one by one,this dissertation creatively regards the non-ideal characteristics as a part of the sensing matrix,and proposes a relative sensing matrix which is convenient to be measured by the known signals.Based on the advantages of wideband modulation signal,which is easy to trigger synchronization and contains the identifiable envelope header,this dissertation implements the measurement of relative sensing matrix.The Normalized Mean Square Error(NMSE)is used to measure the similarity between the reconstructed baseband and the original baseband.The measured relative sensing matrix reduces the NMSE of the reconstructed baseband signal from-7.47 dB to-21.05 dB.3.Based on the high column correlation of the sensing matrix,this dissertation proposes DSOMP algorithm to improve the reconstruction rate of the support set.Due to the limitation of the hardware circuit scale and the influence of the non-ideal characteristics of the devices,the sensing matrix of the compressed sensing receiver fails to strictly meet the Restricted Isometry Property(RIP).Moreover,the sensing matrix with high column correlation will make it difficult for classical greedy algorithms to guide a satisfactory reconstruction rate of the support set.Therefore,this dissertation proposes a DSOMP algorithm,which uses a "wide-in,strict-out" dual-screening mechanism to avoid strong interference atoms blocking the base atoms from entering the candidate set.The distance from the residual to the linear subspace containing selected atoms is used as the selection criterion,so that the selected atom set helps the algorithm to lock the base atoms quickly.In the simulation,with a high-column correlation matrix,the reconstruction rate of the four-carrier signal(at the signal-to-noise ratio of 15 dB)by the DSOMP algorithm is higher than 98%,far exceeding the other greedy algorithms(less than 50%)participating in the comparison.In experiments based on compressed sensing receivers,the successful reconstruction of the support set of the three-carrier signal also proved the practicability of the DSOMP algorithm.At the same time,other greedy algorithms involved in the comparison failed to reconstruct the support set.4.To save the hardware cost of the compressed sensing receiver and reduce the sampling rate required for blind reconstruction,this dissertation proposes an new compressed sampling structure named A-MWC.It realizes that the two MWCs work alternately in the time domain by switching two sets of repeated pseudo-random bit sequences and adding a trigger signal to constrain the timing.Based on the wide stationary signal,the underdetermined equations of two MWCs can be simultaneously established at the power spectrum level to solve the support set,and then the original signal can be reconstructed by the support set.At the cost of a slight amount of calculation,the A-MWC structure reduces the total sampling rate required for signal blind reconstruction to half of the MWC,which also means that the hardware scale is almost reduced by half.Simulations and experiments have proved the theoretical correctness and physical feasibility of the A-MWC.In the experiment,the A-MWC structure with the three sampling branches can successfully reconstruct the four-carrier wideband signal.Compared with the original baseband,the NMSE of the reconstructed four baseband signals are all lower than-15 dB.However,the MWC structure of the same hardware scale failed to reconstruct the support set.