基于谱参数估计的ADC静态测试方法的优化和实现

发布时间：2018-04-19 23:05

本文选题：谱参数估计 + 分段拟合　；参考：《东南大学》2015年硕士论文

【摘要】：近年来,ADC不断的朝着高速高精度的方向发展,这使得ADC标准测试算法的测试成本越来越高,而且采用标准算法测试时需要采集静态和动态两类数据并结合相应的静态和动态算法才能测出ADC所有的参数,这更增加了ADC的测试成本。通过一次数据采集实现动态和静态两类参数的测试将能节约一次数据采集成本,而现有的该种算法有两类,一类是基于静态实现动态参数测试,但该类方法是基于直方图测试实现的,需要采集大量的点数,所减少的成本有限；另一类是基于动态实现静态参数的测试,该类算法是基于拟合和频谱实现的,所需的采样点很少,可以降低大量成本,但该类算法的缺点是测试精度不高。因此,研究能够大量降低测试成本的基于动态实现静态参数测试的单次数据采集实现两类参数测试的算法将具有极大的实用价值。基于此,本文对基于动态实现静态参数测试中的谱参数估计算法进行了研究,并在研究基础上独创性的结合分段理论实现了谱参数估计算法的优化。至此,本文在解决完优化算法实现的关键问题即特定信号源的实现后,该优化算法首先将ADC的传输函数采用傅里叶级数建模为时间与数字代码的表达式,之后通过切比雪夫第一等式建立输出代码与模拟输入之间的关系式,并根据ADC的满幅量程将ADC的传输函数等分成几段,在每段内分别输入正弦、采集数据以及计算傅里叶系数,然后通过连续处理得到ADC完整的传输函数,并与ADC的理想传输函数相减得到ADC的INL曲线,从而实现用动态数据计算ADC的静态参数。此外,本文还重点研究了该优化算法的最佳测试环境。最后,本文以14比特AD9648为基础搭建AD9648的自动化测试系统,通过实验验证本算法的性能。实验结果表明,以直方图算法测试得到的最大INL即1.189LSB为该系统的真实INL,在最佳估算条件即采样点为8000,傅里叶估算项数为80的情况下,原有的谱参数估计算法所得的INL的估算误差达到了0.624LSB,而同样在最佳估算条件即采样点为8000,傅里叶估算项数为8,分段为8的情况下,本文所提算法得到的INL估算误差为0.2953LSB。其估算精度相对原有谱参数估计算法提高了0.3303LSB,近30%。除此之外,本文所提算法比谱参数估计算法少做了近200万次乘法,相比谱参数估计算法结合MATLAB软件所得的计算时间2.605s而言,本文所提算法只需0.2832s,计算时间降低了近90%。
[Abstract]:In recent years, the ADC has been developing in the direction of high speed and high precision, which makes the test cost of ADC standard test algorithm more and more high. Moreover, it is necessary to collect static and dynamic data and combine the corresponding static and dynamic algorithms to measure all the parameters of ADC, which increases the testing cost of ADC. The cost of data acquisition can be saved by realizing the test of dynamic and static parameters through one data acquisition. There are two kinds of algorithms, one is to realize dynamic parameter testing based on static state, the other is to realize dynamic parameter testing based on static state. But this kind of method is based on histogram test, it needs to collect a lot of points, the cost is limited, the other is based on the dynamic implementation of static parameters testing, this kind of algorithm is based on fitting and spectrum implementation. There are few sampling points needed, which can reduce a lot of cost, but the shortcoming of this kind of algorithm is that the test accuracy is not high. Therefore, it is of great practical value to study a single data acquisition algorithm based on dynamic static parameter testing, which can greatly reduce the cost of testing. Based on this, this paper studies the spectral parameter estimation algorithm based on dynamic static parameter testing, and realizes the optimization of spectral parameter estimation algorithm based on the original piecewise theory. So far, after solving the key problem of the realization of the optimization algorithm, that is, the realization of a specific signal source, the optimization algorithm first uses Fourier series to model the transfer function of ADC as the expression of time and digital code. Then the relation between the output code and the analog input is established by Chebyshev's first equation. According to the full amplitude range of ADC, the transfer function of ADC is divided into several sections. In each section, sinusoidal input, data collection and Fourier coefficient are calculated. Then the complete transfer function of ADC is obtained by continuous processing, and the INL curve of ADC is obtained by subtracting it from the ideal transfer function of ADC, so that the static parameters of ADC can be calculated with dynamic data. In addition, this paper also focuses on the optimal test environment of the optimization algorithm. Finally, based on 14 bit AD9648, the automatic test system of AD9648 is built, and the performance of the algorithm is verified by experiments. The experimental results show that the maximum INL obtained by histogram algorithm, that is, 1.189LSB, is the real INL of the system. The best estimation condition is that the sampling point is 8 000 and the Fourier estimation term is 80. The estimation error of INL obtained by the original spectral parameter estimation algorithm is 0.624 LSB. but the estimation error of INL obtained by the proposed algorithm is 0.2953 LSB-0.2953LSB-0.2953LSB-0.2953LSB. when the sample point is 8000, Fourier estimation term is 8 and subsection is 8, the estimation error of INL is 0.2953LSB. Compared with the original spectral parameter estimation algorithm, its estimation accuracy is improved by 0.3303LSBand nearly 30303LSB. In addition, the proposed algorithm is less than the spectral parameter estimation algorithm by nearly 2 million times. Compared with the spectral parameter estimation algorithm combined with the MATLAB software, the computational time of the proposed algorithm is only 0.2832s, and the computational time is reduced by nearly 90 seconds.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN792

【相似文献】