Research On Frequency-domain Characteristics Testing Methods For High Speed And High Precision ADCs

With the rapid development of electronic information technologies, analog-to-digital converters(ADCs), as the digital and analog interface circuits, have been widely used in different areas such as the data acquisition, precision industrial measurement, audio and viedo applications. The frequency-domain characteristics of ADCs are thus receiving more and more attention. The fast Fourier transformation(FFT) is one of the most widely used methods for measuring frequency-domain characteristics, however, it is difficult to achieve coherent sampling and integral period truncation when using the FFT method to test frequency-domain parameters of ADCs. Therefore, the non-coherent sampling is inevitable, and the resultant spectral leakage and picket fence effect(PFE) will decrease the accuracy of measured frequency-domain parameters. In this thesis, the FFT method is mainly used to test the important frequency-domain characteristics of high speed and high precision ADCs, including signal to noise ratio(SNR), signal to noise and distortion(SINAD), effective number of bits(ENOB), spurious-free dynamic range(SFDR) and total harmonic distortion(THD). The approaches based on the average spectrum and the triple-spectrum-line interpolated FFT algorithm are proposed to decrease the errors caused by the spectral leakage and PFE. Main research content of this thesis is introduced as follows.Firstly, the basic structures and working principles of ADCs are introduced with an emphasis on the pipelined ADCs, which are currently the most widely used high speed and high precision ADC structure. The definitions and calculation methods of important frequency-domain parameters of high speed and high precision ADCs are then presented. Meanwhile, their testing methods are explained in detail, including the traditional digital-to-analog converter(DAC) testing method, the code density histogram method, the sine fitting method and the FFT method. The advantages, disadvantages and application ranges of these methods are also compared and summarized.Secondly, Simulink is employed to establish the system-level models for the 12 bits 200 MSPS pipelined ADC and the 14 bits 160 MSPS pipelined ADC. The verification platform is constructed based on the behavior-level dynamic models of AD9230, AD9246 and AD9461, which are typical high-speed ADCs of Analog Devices Incorporation(ADI). Moreover, a suit of hardware testing system for high speed and high precision ADCs is designed and built by using Field Programmable Gate Array(FPGA), high-performance signal source and filter, which can provide simulation and experimental platforms for the proposed testing methods.Thirdly, the average spectrum method is studied in detail. In order to improve the precision of frequency spectrum analysis and reduce the effect of random signals, the output digital signals of ADCs are sampled twice or more, and implemented the windowed FFT algorithms for obtaining the spectrum diagram. The average power spectrum is then obtained by the superposition of the spectrum diagrams. The frequency-domain parameters can finally be calculated from the average spectrum. Both simulation and testing results show that, even in the case of the largest incoherence degree, the measurement results of frequency-domain parameters based on the average spectrum method have obviously smaller errors than those obtained by the windowed FFT method. The results are very close to those obtained under the coherent sampling conditions, satifying the test standards provided by ADI.Finally, in order to reduce the spectral leakage, a cosine window with the maximum side lobe decay speed is built up for signal processing. In order to reduce the error caused by the PFE, the windowed results are corrected by the proposed triple-spectrum-line interpolated FFT algorithm. The amplitude correction formulas based on the frequently used combined cosine window and the cosine window with the maximum side lobe decay speed are obtained by function fitting. The proposed method is compared with the average spectrum method. Both simulation and testing results show that, the average spectrum method can measure parameters such as SNR and ENOB accurately, while the triple-spectrum-line interpolated FFT algorithm is very suitable for testing the SFDR of high speed and high precision ADCs.Two simulation platforms and an experimental testing system are adopted to verify the effectiveness and accuracy of the proposed methods. The results demonstrate that, the proposed methods are universal, and can test frequency-domain characteristics of high speed and high precision ADCs effectively. When these methods are applied to a practical testing system, both the system building difficulty and the test cost can be decreased remarkably.